1f5166768STheodore Ts'o // SPDX-License-Identifier: GPL-2.0 2ac27a0ecSDave Kleikamp /* 3617ba13bSMingming Cao * linux/fs/ext4/super.c 4ac27a0ecSDave Kleikamp * 5ac27a0ecSDave Kleikamp * Copyright (C) 1992, 1993, 1994, 1995 6ac27a0ecSDave Kleikamp * Remy Card (card@masi.ibp.fr) 7ac27a0ecSDave Kleikamp * Laboratoire MASI - Institut Blaise Pascal 8ac27a0ecSDave Kleikamp * Universite Pierre et Marie Curie (Paris VI) 9ac27a0ecSDave Kleikamp * 10ac27a0ecSDave Kleikamp * from 11ac27a0ecSDave Kleikamp * 12ac27a0ecSDave Kleikamp * linux/fs/minix/inode.c 13ac27a0ecSDave Kleikamp * 14ac27a0ecSDave Kleikamp * Copyright (C) 1991, 1992 Linus Torvalds 15ac27a0ecSDave Kleikamp * 16ac27a0ecSDave Kleikamp * Big-endian to little-endian byte-swapping/bitmaps by 17ac27a0ecSDave Kleikamp * David S. Miller (davem@caip.rutgers.edu), 1995 18ac27a0ecSDave Kleikamp */ 19ac27a0ecSDave Kleikamp 20ac27a0ecSDave Kleikamp #include <linux/module.h> 21ac27a0ecSDave Kleikamp #include <linux/string.h> 22ac27a0ecSDave Kleikamp #include <linux/fs.h> 23ac27a0ecSDave Kleikamp #include <linux/time.h> 24c5ca7c76STheodore Ts'o #include <linux/vmalloc.h> 25ac27a0ecSDave Kleikamp #include <linux/slab.h> 26ac27a0ecSDave Kleikamp #include <linux/init.h> 27ac27a0ecSDave Kleikamp #include <linux/blkdev.h> 2866114cadSTejun Heo #include <linux/backing-dev.h> 29ac27a0ecSDave Kleikamp #include <linux/parser.h> 30ac27a0ecSDave Kleikamp #include <linux/buffer_head.h> 31a5694255SChristoph Hellwig #include <linux/exportfs.h> 32ac27a0ecSDave Kleikamp #include <linux/vfs.h> 33ac27a0ecSDave Kleikamp #include <linux/random.h> 34ac27a0ecSDave Kleikamp #include <linux/mount.h> 35ac27a0ecSDave Kleikamp #include <linux/namei.h> 36ac27a0ecSDave Kleikamp #include <linux/quotaops.h> 37ac27a0ecSDave Kleikamp #include <linux/seq_file.h> 383197ebdbSTheodore Ts'o #include <linux/ctype.h> 391330593eSVignesh Babu #include <linux/log2.h> 40717d50e4SAndreas Dilger #include <linux/crc16.h> 41ef510424SDan Williams #include <linux/dax.h> 427abc52c2SDan Magenheimer #include <linux/cleancache.h> 437c0f6ba6SLinus Torvalds #include <linux/uaccess.h> 44ee73f9a5SJeff Layton #include <linux/iversion.h> 45c83ad55eSGabriel Krisman Bertazi #include <linux/unicode.h> 46c6a564ffSChristoph Hellwig #include <linux/part_stat.h> 47bfff6873SLukas Czerner #include <linux/kthread.h> 48bfff6873SLukas Czerner #include <linux/freezer.h> 49bfff6873SLukas Czerner 503dcf5451SChristoph Hellwig #include "ext4.h" 514a092d73STheodore Ts'o #include "ext4_extents.h" /* Needed for trace points definition */ 523dcf5451SChristoph Hellwig #include "ext4_jbd2.h" 53ac27a0ecSDave Kleikamp #include "xattr.h" 54ac27a0ecSDave Kleikamp #include "acl.h" 553661d286STheodore Ts'o #include "mballoc.h" 560c9ec4beSDarrick J. Wong #include "fsmap.h" 57ac27a0ecSDave Kleikamp 589bffad1eSTheodore Ts'o #define CREATE_TRACE_POINTS 599bffad1eSTheodore Ts'o #include <trace/events/ext4.h> 609bffad1eSTheodore Ts'o 610b75a840SLukas Czerner static struct ext4_lazy_init *ext4_li_info; 620b75a840SLukas Czerner static struct mutex ext4_li_mtx; 63e294a537STheodore Ts'o static struct ratelimit_state ext4_mount_msg_ratelimit; 649f6200bbSTheodore Ts'o 65617ba13bSMingming Cao static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 66ac27a0ecSDave Kleikamp unsigned long journal_devnum); 672adf6da8STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root); 68e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync); 6911215630SJan Kara static int ext4_mark_recovery_complete(struct super_block *sb, 70617ba13bSMingming Cao struct ext4_super_block *es); 7111215630SJan Kara static int ext4_clear_journal_err(struct super_block *sb, 72617ba13bSMingming Cao struct ext4_super_block *es); 73617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait); 74617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data); 75617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf); 76c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb); 77c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb); 78152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 79152a0836SAl Viro const char *dev_name, void *data); 802035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb); 812035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb); 82d39195c3SAmir Goldstein static int ext4_feature_set_ok(struct super_block *sb, int readonly); 83bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void); 84bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb); 858f1f7453SEric Sandeen static void ext4_clear_request_list(void); 86c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 87c6cb7e77SEric Whitney unsigned int journal_inum); 88ac27a0ecSDave Kleikamp 89e74031fdSJan Kara /* 90e74031fdSJan Kara * Lock ordering 91e74031fdSJan Kara * 92e74031fdSJan Kara * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and 93e74031fdSJan Kara * i_mmap_rwsem (inode->i_mmap_rwsem)! 94e74031fdSJan Kara * 95e74031fdSJan Kara * page fault path: 96c1e8d7c6SMichel Lespinasse * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start -> 97e74031fdSJan Kara * page lock -> i_data_sem (rw) 98e74031fdSJan Kara * 99e74031fdSJan Kara * buffered write path: 100c1e8d7c6SMichel Lespinasse * sb_start_write -> i_mutex -> mmap_lock 101e74031fdSJan Kara * sb_start_write -> i_mutex -> transaction start -> page lock -> 102e74031fdSJan Kara * i_data_sem (rw) 103e74031fdSJan Kara * 104e74031fdSJan Kara * truncate: 1051d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock 1061d39834fSNikolay Borisov * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start -> 1071d39834fSNikolay Borisov * i_data_sem (rw) 108e74031fdSJan Kara * 109e74031fdSJan Kara * direct IO: 110c1e8d7c6SMichel Lespinasse * sb_start_write -> i_mutex -> mmap_lock 1111d39834fSNikolay Borisov * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw) 112e74031fdSJan Kara * 113e74031fdSJan Kara * writepages: 114e74031fdSJan Kara * transaction start -> page lock(s) -> i_data_sem (rw) 115e74031fdSJan Kara */ 116e74031fdSJan Kara 117c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 1182035e776STheodore Ts'o static struct file_system_type ext2_fs_type = { 1192035e776STheodore Ts'o .owner = THIS_MODULE, 1202035e776STheodore Ts'o .name = "ext2", 1212035e776STheodore Ts'o .mount = ext4_mount, 1222035e776STheodore Ts'o .kill_sb = kill_block_super, 1232035e776STheodore Ts'o .fs_flags = FS_REQUIRES_DEV, 1242035e776STheodore Ts'o }; 1257f78e035SEric W. Biederman MODULE_ALIAS_FS("ext2"); 126fa7614ddSEric W. Biederman MODULE_ALIAS("ext2"); 1272035e776STheodore Ts'o #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type) 1282035e776STheodore Ts'o #else 1292035e776STheodore Ts'o #define IS_EXT2_SB(sb) (0) 1302035e776STheodore Ts'o #endif 1312035e776STheodore Ts'o 1322035e776STheodore Ts'o 133ba69f9abSJan Kara static struct file_system_type ext3_fs_type = { 134ba69f9abSJan Kara .owner = THIS_MODULE, 135ba69f9abSJan Kara .name = "ext3", 136152a0836SAl Viro .mount = ext4_mount, 137ba69f9abSJan Kara .kill_sb = kill_block_super, 138ba69f9abSJan Kara .fs_flags = FS_REQUIRES_DEV, 139ba69f9abSJan Kara }; 1407f78e035SEric W. Biederman MODULE_ALIAS_FS("ext3"); 141fa7614ddSEric W. Biederman MODULE_ALIAS("ext3"); 142ba69f9abSJan Kara #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type) 143bd81d8eeSLaurent Vivier 144fa491b14Szhangyi (F) 145fa491b14Szhangyi (F) static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags, 146fa491b14Szhangyi (F) bh_end_io_t *end_io) 147fa491b14Szhangyi (F) { 148fa491b14Szhangyi (F) /* 149fa491b14Szhangyi (F) * buffer's verified bit is no longer valid after reading from 150fa491b14Szhangyi (F) * disk again due to write out error, clear it to make sure we 151fa491b14Szhangyi (F) * recheck the buffer contents. 152fa491b14Szhangyi (F) */ 153fa491b14Szhangyi (F) clear_buffer_verified(bh); 154fa491b14Szhangyi (F) 155fa491b14Szhangyi (F) bh->b_end_io = end_io ? end_io : end_buffer_read_sync; 156fa491b14Szhangyi (F) get_bh(bh); 157fa491b14Szhangyi (F) submit_bh(REQ_OP_READ, op_flags, bh); 158fa491b14Szhangyi (F) } 159fa491b14Szhangyi (F) 160fa491b14Szhangyi (F) void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags, 161fa491b14Szhangyi (F) bh_end_io_t *end_io) 162fa491b14Szhangyi (F) { 163fa491b14Szhangyi (F) BUG_ON(!buffer_locked(bh)); 164fa491b14Szhangyi (F) 165fa491b14Szhangyi (F) if (ext4_buffer_uptodate(bh)) { 166fa491b14Szhangyi (F) unlock_buffer(bh); 167fa491b14Szhangyi (F) return; 168fa491b14Szhangyi (F) } 169fa491b14Szhangyi (F) __ext4_read_bh(bh, op_flags, end_io); 170fa491b14Szhangyi (F) } 171fa491b14Szhangyi (F) 172fa491b14Szhangyi (F) int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io) 173fa491b14Szhangyi (F) { 174fa491b14Szhangyi (F) BUG_ON(!buffer_locked(bh)); 175fa491b14Szhangyi (F) 176fa491b14Szhangyi (F) if (ext4_buffer_uptodate(bh)) { 177fa491b14Szhangyi (F) unlock_buffer(bh); 178fa491b14Szhangyi (F) return 0; 179fa491b14Szhangyi (F) } 180fa491b14Szhangyi (F) 181fa491b14Szhangyi (F) __ext4_read_bh(bh, op_flags, end_io); 182fa491b14Szhangyi (F) 183fa491b14Szhangyi (F) wait_on_buffer(bh); 184fa491b14Szhangyi (F) if (buffer_uptodate(bh)) 185fa491b14Szhangyi (F) return 0; 186fa491b14Szhangyi (F) return -EIO; 187fa491b14Szhangyi (F) } 188fa491b14Szhangyi (F) 189fa491b14Szhangyi (F) int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait) 190fa491b14Szhangyi (F) { 191fa491b14Szhangyi (F) if (trylock_buffer(bh)) { 192fa491b14Szhangyi (F) if (wait) 193fa491b14Szhangyi (F) return ext4_read_bh(bh, op_flags, NULL); 194fa491b14Szhangyi (F) ext4_read_bh_nowait(bh, op_flags, NULL); 195fa491b14Szhangyi (F) return 0; 196fa491b14Szhangyi (F) } 197fa491b14Szhangyi (F) if (wait) { 198fa491b14Szhangyi (F) wait_on_buffer(bh); 199fa491b14Szhangyi (F) if (buffer_uptodate(bh)) 200fa491b14Szhangyi (F) return 0; 201fa491b14Szhangyi (F) return -EIO; 202fa491b14Szhangyi (F) } 203fa491b14Szhangyi (F) return 0; 204fa491b14Szhangyi (F) } 205fa491b14Szhangyi (F) 206fb265c9cSTheodore Ts'o /* 2078394a6abSzhangyi (F) * This works like __bread_gfp() except it uses ERR_PTR for error 208fb265c9cSTheodore Ts'o * returns. Currently with sb_bread it's impossible to distinguish 209fb265c9cSTheodore Ts'o * between ENOMEM and EIO situations (since both result in a NULL 210fb265c9cSTheodore Ts'o * return. 211fb265c9cSTheodore Ts'o */ 2128394a6abSzhangyi (F) static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb, 2138394a6abSzhangyi (F) sector_t block, int op_flags, 2148394a6abSzhangyi (F) gfp_t gfp) 215fb265c9cSTheodore Ts'o { 2162d069c08Szhangyi (F) struct buffer_head *bh; 2172d069c08Szhangyi (F) int ret; 218fb265c9cSTheodore Ts'o 2198394a6abSzhangyi (F) bh = sb_getblk_gfp(sb, block, gfp); 220fb265c9cSTheodore Ts'o if (bh == NULL) 221fb265c9cSTheodore Ts'o return ERR_PTR(-ENOMEM); 222cf2834a5STheodore Ts'o if (ext4_buffer_uptodate(bh)) 223fb265c9cSTheodore Ts'o return bh; 2242d069c08Szhangyi (F) 2252d069c08Szhangyi (F) ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true); 2262d069c08Szhangyi (F) if (ret) { 227fb265c9cSTheodore Ts'o put_bh(bh); 2282d069c08Szhangyi (F) return ERR_PTR(ret); 2292d069c08Szhangyi (F) } 2302d069c08Szhangyi (F) return bh; 231fb265c9cSTheodore Ts'o } 232fb265c9cSTheodore Ts'o 2338394a6abSzhangyi (F) struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block, 2348394a6abSzhangyi (F) int op_flags) 2358394a6abSzhangyi (F) { 2368394a6abSzhangyi (F) return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE); 2378394a6abSzhangyi (F) } 2388394a6abSzhangyi (F) 2398394a6abSzhangyi (F) struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb, 2408394a6abSzhangyi (F) sector_t block) 2418394a6abSzhangyi (F) { 2428394a6abSzhangyi (F) return __ext4_sb_bread_gfp(sb, block, 0, 0); 2438394a6abSzhangyi (F) } 2448394a6abSzhangyi (F) 2455df1d412Szhangyi (F) void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block) 2465df1d412Szhangyi (F) { 2475df1d412Szhangyi (F) struct buffer_head *bh = sb_getblk_gfp(sb, block, 0); 2485df1d412Szhangyi (F) 2495df1d412Szhangyi (F) if (likely(bh)) { 2505df1d412Szhangyi (F) ext4_read_bh_lock(bh, REQ_RAHEAD, false); 2515df1d412Szhangyi (F) brelse(bh); 2525df1d412Szhangyi (F) } 2535df1d412Szhangyi (F) } 2545df1d412Szhangyi (F) 255d25425f8SDarrick J. Wong static int ext4_verify_csum_type(struct super_block *sb, 256d25425f8SDarrick J. Wong struct ext4_super_block *es) 257d25425f8SDarrick J. Wong { 258e2b911c5SDarrick J. Wong if (!ext4_has_feature_metadata_csum(sb)) 259d25425f8SDarrick J. Wong return 1; 260d25425f8SDarrick J. Wong 261d25425f8SDarrick J. Wong return es->s_checksum_type == EXT4_CRC32C_CHKSUM; 262d25425f8SDarrick J. Wong } 263d25425f8SDarrick J. Wong 264a9c47317SDarrick J. Wong static __le32 ext4_superblock_csum(struct super_block *sb, 265a9c47317SDarrick J. Wong struct ext4_super_block *es) 266a9c47317SDarrick J. Wong { 267a9c47317SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 268a9c47317SDarrick J. Wong int offset = offsetof(struct ext4_super_block, s_checksum); 269a9c47317SDarrick J. Wong __u32 csum; 270a9c47317SDarrick J. Wong 271a9c47317SDarrick J. Wong csum = ext4_chksum(sbi, ~0, (char *)es, offset); 272a9c47317SDarrick J. Wong 273a9c47317SDarrick J. Wong return cpu_to_le32(csum); 274a9c47317SDarrick J. Wong } 275a9c47317SDarrick J. Wong 276c197855eSStephen Hemminger static int ext4_superblock_csum_verify(struct super_block *sb, 277a9c47317SDarrick J. Wong struct ext4_super_block *es) 278a9c47317SDarrick J. Wong { 2799aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 280a9c47317SDarrick J. Wong return 1; 281a9c47317SDarrick J. Wong 282a9c47317SDarrick J. Wong return es->s_checksum == ext4_superblock_csum(sb, es); 283a9c47317SDarrick J. Wong } 284a9c47317SDarrick J. Wong 28506db49e6STheodore Ts'o void ext4_superblock_csum_set(struct super_block *sb) 286a9c47317SDarrick J. Wong { 28706db49e6STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 28806db49e6STheodore Ts'o 2899aa5d32bSDmitry Monakhov if (!ext4_has_metadata_csum(sb)) 290a9c47317SDarrick J. Wong return; 291a9c47317SDarrick J. Wong 292acaa5326SConstantine Sapuntzakis /* 293acaa5326SConstantine Sapuntzakis * Locking the superblock prevents the scenario 294acaa5326SConstantine Sapuntzakis * where: 295acaa5326SConstantine Sapuntzakis * 1) a first thread pauses during checksum calculation. 296acaa5326SConstantine Sapuntzakis * 2) a second thread updates the superblock, recalculates 297acaa5326SConstantine Sapuntzakis * the checksum, and updates s_checksum 298acaa5326SConstantine Sapuntzakis * 3) the first thread resumes and finishes its checksum calculation 299acaa5326SConstantine Sapuntzakis * and updates s_checksum with a potentially stale or torn value. 300acaa5326SConstantine Sapuntzakis */ 301acaa5326SConstantine Sapuntzakis lock_buffer(EXT4_SB(sb)->s_sbh); 302a9c47317SDarrick J. Wong es->s_checksum = ext4_superblock_csum(sb, es); 303acaa5326SConstantine Sapuntzakis unlock_buffer(EXT4_SB(sb)->s_sbh); 304a9c47317SDarrick J. Wong } 305a9c47317SDarrick J. Wong 3068fadc143SAlexandre Ratchov ext4_fsblk_t ext4_block_bitmap(struct super_block *sb, 3078fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 308bd81d8eeSLaurent Vivier { 3093a14589cSAneesh Kumar K.V return le32_to_cpu(bg->bg_block_bitmap_lo) | 3108fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 3118fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0); 312bd81d8eeSLaurent Vivier } 313bd81d8eeSLaurent Vivier 3148fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb, 3158fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 316bd81d8eeSLaurent Vivier { 3175272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_bitmap_lo) | 3188fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 3198fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0); 320bd81d8eeSLaurent Vivier } 321bd81d8eeSLaurent Vivier 3228fadc143SAlexandre Ratchov ext4_fsblk_t ext4_inode_table(struct super_block *sb, 3238fadc143SAlexandre Ratchov struct ext4_group_desc *bg) 324bd81d8eeSLaurent Vivier { 3255272f837SAneesh Kumar K.V return le32_to_cpu(bg->bg_inode_table_lo) | 3268fadc143SAlexandre Ratchov (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 3278fadc143SAlexandre Ratchov (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0); 328bd81d8eeSLaurent Vivier } 329bd81d8eeSLaurent Vivier 330021b65bbSTheodore Ts'o __u32 ext4_free_group_clusters(struct super_block *sb, 331560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 332560671a0SAneesh Kumar K.V { 333560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_blocks_count_lo) | 334560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 335560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0); 336560671a0SAneesh Kumar K.V } 337560671a0SAneesh Kumar K.V 338560671a0SAneesh Kumar K.V __u32 ext4_free_inodes_count(struct super_block *sb, 339560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 340560671a0SAneesh Kumar K.V { 341560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_free_inodes_count_lo) | 342560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 343560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0); 344560671a0SAneesh Kumar K.V } 345560671a0SAneesh Kumar K.V 346560671a0SAneesh Kumar K.V __u32 ext4_used_dirs_count(struct super_block *sb, 347560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 348560671a0SAneesh Kumar K.V { 349560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_used_dirs_count_lo) | 350560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 351560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0); 352560671a0SAneesh Kumar K.V } 353560671a0SAneesh Kumar K.V 354560671a0SAneesh Kumar K.V __u32 ext4_itable_unused_count(struct super_block *sb, 355560671a0SAneesh Kumar K.V struct ext4_group_desc *bg) 356560671a0SAneesh Kumar K.V { 357560671a0SAneesh Kumar K.V return le16_to_cpu(bg->bg_itable_unused_lo) | 358560671a0SAneesh Kumar K.V (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ? 359560671a0SAneesh Kumar K.V (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0); 360560671a0SAneesh Kumar K.V } 361560671a0SAneesh Kumar K.V 3628fadc143SAlexandre Ratchov void ext4_block_bitmap_set(struct super_block *sb, 3638fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 364bd81d8eeSLaurent Vivier { 3653a14589cSAneesh Kumar K.V bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk); 3668fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 3678fadc143SAlexandre Ratchov bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32); 368bd81d8eeSLaurent Vivier } 369bd81d8eeSLaurent Vivier 3708fadc143SAlexandre Ratchov void ext4_inode_bitmap_set(struct super_block *sb, 3718fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 372bd81d8eeSLaurent Vivier { 3735272f837SAneesh Kumar K.V bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk); 3748fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 3758fadc143SAlexandre Ratchov bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32); 376bd81d8eeSLaurent Vivier } 377bd81d8eeSLaurent Vivier 3788fadc143SAlexandre Ratchov void ext4_inode_table_set(struct super_block *sb, 3798fadc143SAlexandre Ratchov struct ext4_group_desc *bg, ext4_fsblk_t blk) 380bd81d8eeSLaurent Vivier { 3815272f837SAneesh Kumar K.V bg->bg_inode_table_lo = cpu_to_le32((u32)blk); 3828fadc143SAlexandre Ratchov if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 3838fadc143SAlexandre Ratchov bg->bg_inode_table_hi = cpu_to_le32(blk >> 32); 384bd81d8eeSLaurent Vivier } 385bd81d8eeSLaurent Vivier 386021b65bbSTheodore Ts'o void ext4_free_group_clusters_set(struct super_block *sb, 387560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 388560671a0SAneesh Kumar K.V { 389560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count); 390560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 391560671a0SAneesh Kumar K.V bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16); 392560671a0SAneesh Kumar K.V } 393560671a0SAneesh Kumar K.V 394560671a0SAneesh Kumar K.V void ext4_free_inodes_set(struct super_block *sb, 395560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 396560671a0SAneesh Kumar K.V { 397560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count); 398560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 399560671a0SAneesh Kumar K.V bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16); 400560671a0SAneesh Kumar K.V } 401560671a0SAneesh Kumar K.V 402560671a0SAneesh Kumar K.V void ext4_used_dirs_set(struct super_block *sb, 403560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 404560671a0SAneesh Kumar K.V { 405560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count); 406560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 407560671a0SAneesh Kumar K.V bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16); 408560671a0SAneesh Kumar K.V } 409560671a0SAneesh Kumar K.V 410560671a0SAneesh Kumar K.V void ext4_itable_unused_set(struct super_block *sb, 411560671a0SAneesh Kumar K.V struct ext4_group_desc *bg, __u32 count) 412560671a0SAneesh Kumar K.V { 413560671a0SAneesh Kumar K.V bg->bg_itable_unused_lo = cpu_to_le16((__u16)count); 414560671a0SAneesh Kumar K.V if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT) 415560671a0SAneesh Kumar K.V bg->bg_itable_unused_hi = cpu_to_le16(count >> 16); 416560671a0SAneesh Kumar K.V } 417560671a0SAneesh Kumar K.V 4186a0678a7SArnd Bergmann static void __ext4_update_tstamp(__le32 *lo, __u8 *hi) 4196a0678a7SArnd Bergmann { 4206a0678a7SArnd Bergmann time64_t now = ktime_get_real_seconds(); 4216a0678a7SArnd Bergmann 4226a0678a7SArnd Bergmann now = clamp_val(now, 0, (1ull << 40) - 1); 4236a0678a7SArnd Bergmann 4246a0678a7SArnd Bergmann *lo = cpu_to_le32(lower_32_bits(now)); 4256a0678a7SArnd Bergmann *hi = upper_32_bits(now); 4266a0678a7SArnd Bergmann } 4276a0678a7SArnd Bergmann 4286a0678a7SArnd Bergmann static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi) 4296a0678a7SArnd Bergmann { 4306a0678a7SArnd Bergmann return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo); 4316a0678a7SArnd Bergmann } 4326a0678a7SArnd Bergmann #define ext4_update_tstamp(es, tstamp) \ 4336a0678a7SArnd Bergmann __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 4346a0678a7SArnd Bergmann #define ext4_get_tstamp(es, tstamp) \ 4356a0678a7SArnd Bergmann __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi) 436d3d1faf6SCurt Wohlgemuth 43754d3adbcSTheodore Ts'o static void __save_error_info(struct super_block *sb, int error, 43854d3adbcSTheodore Ts'o __u32 ino, __u64 block, 43954d3adbcSTheodore Ts'o const char *func, unsigned int line) 4401c13d5c0STheodore Ts'o { 4411c13d5c0STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 44254d3adbcSTheodore Ts'o int err; 4431c13d5c0STheodore Ts'o 4441c13d5c0STheodore Ts'o EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 4451b46617bSTheodore Ts'o if (bdev_read_only(sb->s_bdev)) 4461b46617bSTheodore Ts'o return; 4471c13d5c0STheodore Ts'o es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 4486a0678a7SArnd Bergmann ext4_update_tstamp(es, s_last_error_time); 4491c13d5c0STheodore Ts'o strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func)); 4501c13d5c0STheodore Ts'o es->s_last_error_line = cpu_to_le32(line); 45154d3adbcSTheodore Ts'o es->s_last_error_ino = cpu_to_le32(ino); 45254d3adbcSTheodore Ts'o es->s_last_error_block = cpu_to_le64(block); 45354d3adbcSTheodore Ts'o switch (error) { 45454d3adbcSTheodore Ts'o case EIO: 45554d3adbcSTheodore Ts'o err = EXT4_ERR_EIO; 45654d3adbcSTheodore Ts'o break; 45754d3adbcSTheodore Ts'o case ENOMEM: 45854d3adbcSTheodore Ts'o err = EXT4_ERR_ENOMEM; 45954d3adbcSTheodore Ts'o break; 46054d3adbcSTheodore Ts'o case EFSBADCRC: 46154d3adbcSTheodore Ts'o err = EXT4_ERR_EFSBADCRC; 46254d3adbcSTheodore Ts'o break; 46354d3adbcSTheodore Ts'o case 0: 46454d3adbcSTheodore Ts'o case EFSCORRUPTED: 46554d3adbcSTheodore Ts'o err = EXT4_ERR_EFSCORRUPTED; 46654d3adbcSTheodore Ts'o break; 46754d3adbcSTheodore Ts'o case ENOSPC: 46854d3adbcSTheodore Ts'o err = EXT4_ERR_ENOSPC; 46954d3adbcSTheodore Ts'o break; 47054d3adbcSTheodore Ts'o case ENOKEY: 47154d3adbcSTheodore Ts'o err = EXT4_ERR_ENOKEY; 47254d3adbcSTheodore Ts'o break; 47354d3adbcSTheodore Ts'o case EROFS: 47454d3adbcSTheodore Ts'o err = EXT4_ERR_EROFS; 47554d3adbcSTheodore Ts'o break; 47654d3adbcSTheodore Ts'o case EFBIG: 47754d3adbcSTheodore Ts'o err = EXT4_ERR_EFBIG; 47854d3adbcSTheodore Ts'o break; 47954d3adbcSTheodore Ts'o case EEXIST: 48054d3adbcSTheodore Ts'o err = EXT4_ERR_EEXIST; 48154d3adbcSTheodore Ts'o break; 48254d3adbcSTheodore Ts'o case ERANGE: 48354d3adbcSTheodore Ts'o err = EXT4_ERR_ERANGE; 48454d3adbcSTheodore Ts'o break; 48554d3adbcSTheodore Ts'o case EOVERFLOW: 48654d3adbcSTheodore Ts'o err = EXT4_ERR_EOVERFLOW; 48754d3adbcSTheodore Ts'o break; 48854d3adbcSTheodore Ts'o case EBUSY: 48954d3adbcSTheodore Ts'o err = EXT4_ERR_EBUSY; 49054d3adbcSTheodore Ts'o break; 49154d3adbcSTheodore Ts'o case ENOTDIR: 49254d3adbcSTheodore Ts'o err = EXT4_ERR_ENOTDIR; 49354d3adbcSTheodore Ts'o break; 49454d3adbcSTheodore Ts'o case ENOTEMPTY: 49554d3adbcSTheodore Ts'o err = EXT4_ERR_ENOTEMPTY; 49654d3adbcSTheodore Ts'o break; 49754d3adbcSTheodore Ts'o case ESHUTDOWN: 49854d3adbcSTheodore Ts'o err = EXT4_ERR_ESHUTDOWN; 49954d3adbcSTheodore Ts'o break; 50054d3adbcSTheodore Ts'o case EFAULT: 50154d3adbcSTheodore Ts'o err = EXT4_ERR_EFAULT; 50254d3adbcSTheodore Ts'o break; 50354d3adbcSTheodore Ts'o default: 50454d3adbcSTheodore Ts'o err = EXT4_ERR_UNKNOWN; 50554d3adbcSTheodore Ts'o } 50654d3adbcSTheodore Ts'o es->s_last_error_errcode = err; 5071c13d5c0STheodore Ts'o if (!es->s_first_error_time) { 5081c13d5c0STheodore Ts'o es->s_first_error_time = es->s_last_error_time; 5096a0678a7SArnd Bergmann es->s_first_error_time_hi = es->s_last_error_time_hi; 5101c13d5c0STheodore Ts'o strncpy(es->s_first_error_func, func, 5111c13d5c0STheodore Ts'o sizeof(es->s_first_error_func)); 5121c13d5c0STheodore Ts'o es->s_first_error_line = cpu_to_le32(line); 5131c13d5c0STheodore Ts'o es->s_first_error_ino = es->s_last_error_ino; 5141c13d5c0STheodore Ts'o es->s_first_error_block = es->s_last_error_block; 515878520acSTheodore Ts'o es->s_first_error_errcode = es->s_last_error_errcode; 5161c13d5c0STheodore Ts'o } 51766e61a9eSTheodore Ts'o /* 51866e61a9eSTheodore Ts'o * Start the daily error reporting function if it hasn't been 51966e61a9eSTheodore Ts'o * started already 52066e61a9eSTheodore Ts'o */ 52166e61a9eSTheodore Ts'o if (!es->s_error_count) 52266e61a9eSTheodore Ts'o mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ); 523ba39ebb6SWei Yongjun le32_add_cpu(&es->s_error_count, 1); 5241c13d5c0STheodore Ts'o } 5251c13d5c0STheodore Ts'o 52654d3adbcSTheodore Ts'o static void save_error_info(struct super_block *sb, int error, 52754d3adbcSTheodore Ts'o __u32 ino, __u64 block, 52854d3adbcSTheodore Ts'o const char *func, unsigned int line) 5291c13d5c0STheodore Ts'o { 53054d3adbcSTheodore Ts'o __save_error_info(sb, error, ino, block, func, line); 531c96e2b85SEric Sandeen if (!bdev_read_only(sb->s_bdev)) 5321c13d5c0STheodore Ts'o ext4_commit_super(sb, 1); 5331c13d5c0STheodore Ts'o } 5341c13d5c0STheodore Ts'o 535bdfe0cbdSTheodore Ts'o /* 536bdfe0cbdSTheodore Ts'o * The del_gendisk() function uninitializes the disk-specific data 537bdfe0cbdSTheodore Ts'o * structures, including the bdi structure, without telling anyone 538bdfe0cbdSTheodore Ts'o * else. Once this happens, any attempt to call mark_buffer_dirty() 539bdfe0cbdSTheodore Ts'o * (for example, by ext4_commit_super), will cause a kernel OOPS. 540bdfe0cbdSTheodore Ts'o * This is a kludge to prevent these oops until we can put in a proper 541bdfe0cbdSTheodore Ts'o * hook in del_gendisk() to inform the VFS and file system layers. 542bdfe0cbdSTheodore Ts'o */ 543bdfe0cbdSTheodore Ts'o static int block_device_ejected(struct super_block *sb) 544bdfe0cbdSTheodore Ts'o { 545bdfe0cbdSTheodore Ts'o struct inode *bd_inode = sb->s_bdev->bd_inode; 546bdfe0cbdSTheodore Ts'o struct backing_dev_info *bdi = inode_to_bdi(bd_inode); 547bdfe0cbdSTheodore Ts'o 548bdfe0cbdSTheodore Ts'o return bdi->dev == NULL; 549bdfe0cbdSTheodore Ts'o } 550bdfe0cbdSTheodore Ts'o 55118aadd47SBobi Jam static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn) 55218aadd47SBobi Jam { 55318aadd47SBobi Jam struct super_block *sb = journal->j_private; 55418aadd47SBobi Jam struct ext4_sb_info *sbi = EXT4_SB(sb); 55518aadd47SBobi Jam int error = is_journal_aborted(journal); 5565d3ee208SDmitry Monakhov struct ext4_journal_cb_entry *jce; 55718aadd47SBobi Jam 5585d3ee208SDmitry Monakhov BUG_ON(txn->t_state == T_FINISHED); 559a0154344SDaeho Jeong 560a0154344SDaeho Jeong ext4_process_freed_data(sb, txn->t_tid); 561a0154344SDaeho Jeong 56218aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 5635d3ee208SDmitry Monakhov while (!list_empty(&txn->t_private_list)) { 5645d3ee208SDmitry Monakhov jce = list_entry(txn->t_private_list.next, 5655d3ee208SDmitry Monakhov struct ext4_journal_cb_entry, jce_list); 56618aadd47SBobi Jam list_del_init(&jce->jce_list); 56718aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 56818aadd47SBobi Jam jce->jce_func(sb, jce, error); 56918aadd47SBobi Jam spin_lock(&sbi->s_md_lock); 57018aadd47SBobi Jam } 57118aadd47SBobi Jam spin_unlock(&sbi->s_md_lock); 57218aadd47SBobi Jam } 5731c13d5c0STheodore Ts'o 5741dc1097fSJan Kara static bool system_going_down(void) 5751dc1097fSJan Kara { 5761dc1097fSJan Kara return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF 5771dc1097fSJan Kara || system_state == SYSTEM_RESTART; 5781dc1097fSJan Kara } 5791dc1097fSJan Kara 580ac27a0ecSDave Kleikamp /* Deal with the reporting of failure conditions on a filesystem such as 581ac27a0ecSDave Kleikamp * inconsistencies detected or read IO failures. 582ac27a0ecSDave Kleikamp * 583ac27a0ecSDave Kleikamp * On ext2, we can store the error state of the filesystem in the 584617ba13bSMingming Cao * superblock. That is not possible on ext4, because we may have other 585ac27a0ecSDave Kleikamp * write ordering constraints on the superblock which prevent us from 586ac27a0ecSDave Kleikamp * writing it out straight away; and given that the journal is about to 587ac27a0ecSDave Kleikamp * be aborted, we can't rely on the current, or future, transactions to 588ac27a0ecSDave Kleikamp * write out the superblock safely. 589ac27a0ecSDave Kleikamp * 590dab291afSMingming Cao * We'll just use the jbd2_journal_abort() error code to record an error in 591d6b198bcSThadeu Lima de Souza Cascardo * the journal instead. On recovery, the journal will complain about 592ac27a0ecSDave Kleikamp * that error until we've noted it down and cleared it. 593ac27a0ecSDave Kleikamp */ 594ac27a0ecSDave Kleikamp 595617ba13bSMingming Cao static void ext4_handle_error(struct super_block *sb) 596ac27a0ecSDave Kleikamp { 597327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 598327eaf73STheodore Ts'o WARN_ON_ONCE(1); 599327eaf73STheodore Ts'o 600bc98a42cSDavid Howells if (sb_rdonly(sb)) 601ac27a0ecSDave Kleikamp return; 602ac27a0ecSDave Kleikamp 603ac27a0ecSDave Kleikamp if (!test_opt(sb, ERRORS_CONT)) { 604617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 605ac27a0ecSDave Kleikamp 6064ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 607ac27a0ecSDave Kleikamp if (journal) 608dab291afSMingming Cao jbd2_journal_abort(journal, -EIO); 609ac27a0ecSDave Kleikamp } 6101dc1097fSJan Kara /* 6111dc1097fSJan Kara * We force ERRORS_RO behavior when system is rebooting. Otherwise we 6121dc1097fSJan Kara * could panic during 'reboot -f' as the underlying device got already 6131dc1097fSJan Kara * disabled. 6141dc1097fSJan Kara */ 6151dc1097fSJan Kara if (test_opt(sb, ERRORS_RO) || system_going_down()) { 616b31e1552SEric Sandeen ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 6174418e141SDmitry Monakhov /* 6184418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 6194418e141SDmitry Monakhov * before ->s_flags update 6204418e141SDmitry Monakhov */ 6214418e141SDmitry Monakhov smp_wmb(); 6221751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 6231dc1097fSJan Kara } else if (test_opt(sb, ERRORS_PANIC)) { 624617ba13bSMingming Cao panic("EXT4-fs (device %s): panic forced after error\n", 625ac27a0ecSDave Kleikamp sb->s_id); 626ac27a0ecSDave Kleikamp } 6274327ba52SDaeho Jeong } 628ac27a0ecSDave Kleikamp 629efbed4dcSTheodore Ts'o #define ext4_error_ratelimit(sb) \ 630efbed4dcSTheodore Ts'o ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \ 631efbed4dcSTheodore Ts'o "EXT4-fs error") 632efbed4dcSTheodore Ts'o 63312062dddSEric Sandeen void __ext4_error(struct super_block *sb, const char *function, 63454d3adbcSTheodore Ts'o unsigned int line, int error, __u64 block, 63554d3adbcSTheodore Ts'o const char *fmt, ...) 636ac27a0ecSDave Kleikamp { 6370ff2ea7dSJoe Perches struct va_format vaf; 638ac27a0ecSDave Kleikamp va_list args; 639ac27a0ecSDave Kleikamp 6400db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 6410db1ff22STheodore Ts'o return; 6420db1ff22STheodore Ts'o 643ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 644efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 645ac27a0ecSDave Kleikamp va_start(args, fmt); 6460ff2ea7dSJoe Perches vaf.fmt = fmt; 6470ff2ea7dSJoe Perches vaf.va = &args; 648efbed4dcSTheodore Ts'o printk(KERN_CRIT 649efbed4dcSTheodore Ts'o "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n", 6500ff2ea7dSJoe Perches sb->s_id, function, line, current->comm, &vaf); 651ac27a0ecSDave Kleikamp va_end(args); 652efbed4dcSTheodore Ts'o } 65354d3adbcSTheodore Ts'o save_error_info(sb, error, 0, block, function, line); 654617ba13bSMingming Cao ext4_handle_error(sb); 655ac27a0ecSDave Kleikamp } 656ac27a0ecSDave Kleikamp 657e7c96e8eSJoe Perches void __ext4_error_inode(struct inode *inode, const char *function, 65854d3adbcSTheodore Ts'o unsigned int line, ext4_fsblk_t block, int error, 659273df556SFrank Mayhar const char *fmt, ...) 660273df556SFrank Mayhar { 661273df556SFrank Mayhar va_list args; 662f7c21177STheodore Ts'o struct va_format vaf; 663273df556SFrank Mayhar 6640db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 6650db1ff22STheodore Ts'o return; 6660db1ff22STheodore Ts'o 667ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 668efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 669273df556SFrank Mayhar va_start(args, fmt); 670f7c21177STheodore Ts'o vaf.fmt = fmt; 671f7c21177STheodore Ts'o vaf.va = &args; 672c398eda0STheodore Ts'o if (block) 673d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 674d9ee81daSJoe Perches "inode #%lu: block %llu: comm %s: %pV\n", 675d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 676d9ee81daSJoe Perches block, current->comm, &vaf); 677d9ee81daSJoe Perches else 678d9ee81daSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: " 679d9ee81daSJoe Perches "inode #%lu: comm %s: %pV\n", 680d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 681d9ee81daSJoe Perches current->comm, &vaf); 682273df556SFrank Mayhar va_end(args); 683efbed4dcSTheodore Ts'o } 68454d3adbcSTheodore Ts'o save_error_info(inode->i_sb, error, inode->i_ino, block, 68554d3adbcSTheodore Ts'o function, line); 686273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 687273df556SFrank Mayhar } 688273df556SFrank Mayhar 689e7c96e8eSJoe Perches void __ext4_error_file(struct file *file, const char *function, 690f7c21177STheodore Ts'o unsigned int line, ext4_fsblk_t block, 691f7c21177STheodore Ts'o const char *fmt, ...) 692273df556SFrank Mayhar { 693273df556SFrank Mayhar va_list args; 694f7c21177STheodore Ts'o struct va_format vaf; 695496ad9aaSAl Viro struct inode *inode = file_inode(file); 696273df556SFrank Mayhar char pathname[80], *path; 697273df556SFrank Mayhar 6980db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb)))) 6990db1ff22STheodore Ts'o return; 7000db1ff22STheodore Ts'o 701ccf0f32aSTheodore Ts'o trace_ext4_error(inode->i_sb, function, line); 702efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(inode->i_sb)) { 7039bf39ab2SMiklos Szeredi path = file_path(file, pathname, sizeof(pathname)); 704f9a62d09SDan Carpenter if (IS_ERR(path)) 705273df556SFrank Mayhar path = "(unknown)"; 706f7c21177STheodore Ts'o va_start(args, fmt); 707f7c21177STheodore Ts'o vaf.fmt = fmt; 708f7c21177STheodore Ts'o vaf.va = &args; 709d9ee81daSJoe Perches if (block) 710d9ee81daSJoe Perches printk(KERN_CRIT 711d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 712d9ee81daSJoe Perches "block %llu: comm %s: path %s: %pV\n", 713d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 714d9ee81daSJoe Perches block, current->comm, path, &vaf); 715d9ee81daSJoe Perches else 716d9ee81daSJoe Perches printk(KERN_CRIT 717d9ee81daSJoe Perches "EXT4-fs error (device %s): %s:%d: inode #%lu: " 718d9ee81daSJoe Perches "comm %s: path %s: %pV\n", 719d9ee81daSJoe Perches inode->i_sb->s_id, function, line, inode->i_ino, 720d9ee81daSJoe Perches current->comm, path, &vaf); 721273df556SFrank Mayhar va_end(args); 722efbed4dcSTheodore Ts'o } 72354d3adbcSTheodore Ts'o save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block, 72454d3adbcSTheodore Ts'o function, line); 725273df556SFrank Mayhar ext4_handle_error(inode->i_sb); 726273df556SFrank Mayhar } 727273df556SFrank Mayhar 728722887ddSTheodore Ts'o const char *ext4_decode_error(struct super_block *sb, int errno, 729ac27a0ecSDave Kleikamp char nbuf[16]) 730ac27a0ecSDave Kleikamp { 731ac27a0ecSDave Kleikamp char *errstr = NULL; 732ac27a0ecSDave Kleikamp 733ac27a0ecSDave Kleikamp switch (errno) { 7346a797d27SDarrick J. Wong case -EFSCORRUPTED: 7356a797d27SDarrick J. Wong errstr = "Corrupt filesystem"; 7366a797d27SDarrick J. Wong break; 7376a797d27SDarrick J. Wong case -EFSBADCRC: 7386a797d27SDarrick J. Wong errstr = "Filesystem failed CRC"; 7396a797d27SDarrick J. Wong break; 740ac27a0ecSDave Kleikamp case -EIO: 741ac27a0ecSDave Kleikamp errstr = "IO failure"; 742ac27a0ecSDave Kleikamp break; 743ac27a0ecSDave Kleikamp case -ENOMEM: 744ac27a0ecSDave Kleikamp errstr = "Out of memory"; 745ac27a0ecSDave Kleikamp break; 746ac27a0ecSDave Kleikamp case -EROFS: 74778f1ddbbSTheodore Ts'o if (!sb || (EXT4_SB(sb)->s_journal && 74878f1ddbbSTheodore Ts'o EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)) 749ac27a0ecSDave Kleikamp errstr = "Journal has aborted"; 750ac27a0ecSDave Kleikamp else 751ac27a0ecSDave Kleikamp errstr = "Readonly filesystem"; 752ac27a0ecSDave Kleikamp break; 753ac27a0ecSDave Kleikamp default: 754ac27a0ecSDave Kleikamp /* If the caller passed in an extra buffer for unknown 755ac27a0ecSDave Kleikamp * errors, textualise them now. Else we just return 756ac27a0ecSDave Kleikamp * NULL. */ 757ac27a0ecSDave Kleikamp if (nbuf) { 758ac27a0ecSDave Kleikamp /* Check for truncated error codes... */ 759ac27a0ecSDave Kleikamp if (snprintf(nbuf, 16, "error %d", -errno) >= 0) 760ac27a0ecSDave Kleikamp errstr = nbuf; 761ac27a0ecSDave Kleikamp } 762ac27a0ecSDave Kleikamp break; 763ac27a0ecSDave Kleikamp } 764ac27a0ecSDave Kleikamp 765ac27a0ecSDave Kleikamp return errstr; 766ac27a0ecSDave Kleikamp } 767ac27a0ecSDave Kleikamp 768617ba13bSMingming Cao /* __ext4_std_error decodes expected errors from journaling functions 769ac27a0ecSDave Kleikamp * automatically and invokes the appropriate error response. */ 770ac27a0ecSDave Kleikamp 771c398eda0STheodore Ts'o void __ext4_std_error(struct super_block *sb, const char *function, 772c398eda0STheodore Ts'o unsigned int line, int errno) 773ac27a0ecSDave Kleikamp { 774ac27a0ecSDave Kleikamp char nbuf[16]; 775ac27a0ecSDave Kleikamp const char *errstr; 776ac27a0ecSDave Kleikamp 7770db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 7780db1ff22STheodore Ts'o return; 7790db1ff22STheodore Ts'o 780ac27a0ecSDave Kleikamp /* Special case: if the error is EROFS, and we're not already 781ac27a0ecSDave Kleikamp * inside a transaction, then there's really no point in logging 782ac27a0ecSDave Kleikamp * an error. */ 783bc98a42cSDavid Howells if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb)) 784ac27a0ecSDave Kleikamp return; 785ac27a0ecSDave Kleikamp 786efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 787617ba13bSMingming Cao errstr = ext4_decode_error(sb, errno, nbuf); 788c398eda0STheodore Ts'o printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n", 789c398eda0STheodore Ts'o sb->s_id, function, line, errstr); 790efbed4dcSTheodore Ts'o } 791ac27a0ecSDave Kleikamp 79254d3adbcSTheodore Ts'o save_error_info(sb, -errno, 0, 0, function, line); 793617ba13bSMingming Cao ext4_handle_error(sb); 794ac27a0ecSDave Kleikamp } 795ac27a0ecSDave Kleikamp 796ac27a0ecSDave Kleikamp /* 797617ba13bSMingming Cao * ext4_abort is a much stronger failure handler than ext4_error. The 798ac27a0ecSDave Kleikamp * abort function may be used to deal with unrecoverable failures such 799ac27a0ecSDave Kleikamp * as journal IO errors or ENOMEM at a critical moment in log management. 800ac27a0ecSDave Kleikamp * 801ac27a0ecSDave Kleikamp * We unconditionally force the filesystem into an ABORT|READONLY state, 802ac27a0ecSDave Kleikamp * unless the error response on the fs has been set to panic in which 803ac27a0ecSDave Kleikamp * case we take the easy way out and panic immediately. 804ac27a0ecSDave Kleikamp */ 805ac27a0ecSDave Kleikamp 806c67d859eSTheodore Ts'o void __ext4_abort(struct super_block *sb, const char *function, 80754d3adbcSTheodore Ts'o unsigned int line, int error, const char *fmt, ...) 808ac27a0ecSDave Kleikamp { 809651e1c3bSJoe Perches struct va_format vaf; 810ac27a0ecSDave Kleikamp va_list args; 811ac27a0ecSDave Kleikamp 8120db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 8130db1ff22STheodore Ts'o return; 8140db1ff22STheodore Ts'o 81554d3adbcSTheodore Ts'o save_error_info(sb, error, 0, 0, function, line); 816ac27a0ecSDave Kleikamp va_start(args, fmt); 817651e1c3bSJoe Perches vaf.fmt = fmt; 818651e1c3bSJoe Perches vaf.va = &args; 819651e1c3bSJoe Perches printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n", 820651e1c3bSJoe Perches sb->s_id, function, line, &vaf); 821ac27a0ecSDave Kleikamp va_end(args); 822ac27a0ecSDave Kleikamp 823bc98a42cSDavid Howells if (sb_rdonly(sb) == 0) { 8244ab2f15bSTheodore Ts'o EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED; 8257b97d868Szhangyi (F) if (EXT4_SB(sb)->s_journal) 8267b97d868Szhangyi (F) jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO); 8277b97d868Szhangyi (F) 8287b97d868Szhangyi (F) ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 8294418e141SDmitry Monakhov /* 8304418e141SDmitry Monakhov * Make sure updated value of ->s_mount_flags will be visible 8314418e141SDmitry Monakhov * before ->s_flags update 8324418e141SDmitry Monakhov */ 8334418e141SDmitry Monakhov smp_wmb(); 8341751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 8351c13d5c0STheodore Ts'o } 8367b97d868Szhangyi (F) if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) 8371c13d5c0STheodore Ts'o panic("EXT4-fs panic from previous error\n"); 838ac27a0ecSDave Kleikamp } 839ac27a0ecSDave Kleikamp 840e7c96e8eSJoe Perches void __ext4_msg(struct super_block *sb, 841e7c96e8eSJoe Perches const char *prefix, const char *fmt, ...) 842b31e1552SEric Sandeen { 8430ff2ea7dSJoe Perches struct va_format vaf; 844b31e1552SEric Sandeen va_list args; 845b31e1552SEric Sandeen 8461cf006edSDmitry Monakhov atomic_inc(&EXT4_SB(sb)->s_msg_count); 847efbed4dcSTheodore Ts'o if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs")) 848efbed4dcSTheodore Ts'o return; 849efbed4dcSTheodore Ts'o 850b31e1552SEric Sandeen va_start(args, fmt); 8510ff2ea7dSJoe Perches vaf.fmt = fmt; 8520ff2ea7dSJoe Perches vaf.va = &args; 8530ff2ea7dSJoe Perches printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf); 854b31e1552SEric Sandeen va_end(args); 855b31e1552SEric Sandeen } 856b31e1552SEric Sandeen 8571cf006edSDmitry Monakhov static int ext4_warning_ratelimit(struct super_block *sb) 8581cf006edSDmitry Monakhov { 8591cf006edSDmitry Monakhov atomic_inc(&EXT4_SB(sb)->s_warning_count); 8601cf006edSDmitry Monakhov return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), 8611cf006edSDmitry Monakhov "EXT4-fs warning"); 8621cf006edSDmitry Monakhov } 863b03a2f7eSAndreas Dilger 86412062dddSEric Sandeen void __ext4_warning(struct super_block *sb, const char *function, 865c398eda0STheodore Ts'o unsigned int line, const char *fmt, ...) 866ac27a0ecSDave Kleikamp { 8670ff2ea7dSJoe Perches struct va_format vaf; 868ac27a0ecSDave Kleikamp va_list args; 869ac27a0ecSDave Kleikamp 870b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(sb)) 871efbed4dcSTheodore Ts'o return; 872efbed4dcSTheodore Ts'o 873ac27a0ecSDave Kleikamp va_start(args, fmt); 8740ff2ea7dSJoe Perches vaf.fmt = fmt; 8750ff2ea7dSJoe Perches vaf.va = &args; 8760ff2ea7dSJoe Perches printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n", 8770ff2ea7dSJoe Perches sb->s_id, function, line, &vaf); 878ac27a0ecSDave Kleikamp va_end(args); 879ac27a0ecSDave Kleikamp } 880ac27a0ecSDave Kleikamp 881b03a2f7eSAndreas Dilger void __ext4_warning_inode(const struct inode *inode, const char *function, 882b03a2f7eSAndreas Dilger unsigned int line, const char *fmt, ...) 883b03a2f7eSAndreas Dilger { 884b03a2f7eSAndreas Dilger struct va_format vaf; 885b03a2f7eSAndreas Dilger va_list args; 886b03a2f7eSAndreas Dilger 887b03a2f7eSAndreas Dilger if (!ext4_warning_ratelimit(inode->i_sb)) 888b03a2f7eSAndreas Dilger return; 889b03a2f7eSAndreas Dilger 890b03a2f7eSAndreas Dilger va_start(args, fmt); 891b03a2f7eSAndreas Dilger vaf.fmt = fmt; 892b03a2f7eSAndreas Dilger vaf.va = &args; 893b03a2f7eSAndreas Dilger printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: " 894b03a2f7eSAndreas Dilger "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id, 895b03a2f7eSAndreas Dilger function, line, inode->i_ino, current->comm, &vaf); 896b03a2f7eSAndreas Dilger va_end(args); 897b03a2f7eSAndreas Dilger } 898b03a2f7eSAndreas Dilger 899e29136f8STheodore Ts'o void __ext4_grp_locked_error(const char *function, unsigned int line, 900e29136f8STheodore Ts'o struct super_block *sb, ext4_group_t grp, 901e29136f8STheodore Ts'o unsigned long ino, ext4_fsblk_t block, 902e29136f8STheodore Ts'o const char *fmt, ...) 9035d1b1b3fSAneesh Kumar K.V __releases(bitlock) 9045d1b1b3fSAneesh Kumar K.V __acquires(bitlock) 9055d1b1b3fSAneesh Kumar K.V { 9060ff2ea7dSJoe Perches struct va_format vaf; 9075d1b1b3fSAneesh Kumar K.V va_list args; 9085d1b1b3fSAneesh Kumar K.V 9090db1ff22STheodore Ts'o if (unlikely(ext4_forced_shutdown(EXT4_SB(sb)))) 9100db1ff22STheodore Ts'o return; 9110db1ff22STheodore Ts'o 912ccf0f32aSTheodore Ts'o trace_ext4_error(sb, function, line); 91354d3adbcSTheodore Ts'o __save_error_info(sb, EFSCORRUPTED, ino, block, function, line); 9140ff2ea7dSJoe Perches 915efbed4dcSTheodore Ts'o if (ext4_error_ratelimit(sb)) { 9165d1b1b3fSAneesh Kumar K.V va_start(args, fmt); 9170ff2ea7dSJoe Perches vaf.fmt = fmt; 9180ff2ea7dSJoe Perches vaf.va = &args; 91921149d61SRobin Dong printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ", 920e29136f8STheodore Ts'o sb->s_id, function, line, grp); 921e29136f8STheodore Ts'o if (ino) 9220ff2ea7dSJoe Perches printk(KERN_CONT "inode %lu: ", ino); 923e29136f8STheodore Ts'o if (block) 924efbed4dcSTheodore Ts'o printk(KERN_CONT "block %llu:", 925efbed4dcSTheodore Ts'o (unsigned long long) block); 9260ff2ea7dSJoe Perches printk(KERN_CONT "%pV\n", &vaf); 9275d1b1b3fSAneesh Kumar K.V va_end(args); 928efbed4dcSTheodore Ts'o } 9295d1b1b3fSAneesh Kumar K.V 930327eaf73STheodore Ts'o if (test_opt(sb, WARN_ON_ERROR)) 931327eaf73STheodore Ts'o WARN_ON_ONCE(1); 932327eaf73STheodore Ts'o 9335d1b1b3fSAneesh Kumar K.V if (test_opt(sb, ERRORS_CONT)) { 934e2d67052STheodore Ts'o ext4_commit_super(sb, 0); 9355d1b1b3fSAneesh Kumar K.V return; 9365d1b1b3fSAneesh Kumar K.V } 9371c13d5c0STheodore Ts'o 9385d1b1b3fSAneesh Kumar K.V ext4_unlock_group(sb, grp); 93906f29cc8SZhouyi Zhou ext4_commit_super(sb, 1); 9405d1b1b3fSAneesh Kumar K.V ext4_handle_error(sb); 9415d1b1b3fSAneesh Kumar K.V /* 9425d1b1b3fSAneesh Kumar K.V * We only get here in the ERRORS_RO case; relocking the group 9435d1b1b3fSAneesh Kumar K.V * may be dangerous, but nothing bad will happen since the 9445d1b1b3fSAneesh Kumar K.V * filesystem will have already been marked read/only and the 9455d1b1b3fSAneesh Kumar K.V * journal has been aborted. We return 1 as a hint to callers 9465d1b1b3fSAneesh Kumar K.V * who might what to use the return value from 94725985edcSLucas De Marchi * ext4_grp_locked_error() to distinguish between the 9485d1b1b3fSAneesh Kumar K.V * ERRORS_CONT and ERRORS_RO case, and perhaps return more 9495d1b1b3fSAneesh Kumar K.V * aggressively from the ext4 function in question, with a 9505d1b1b3fSAneesh Kumar K.V * more appropriate error code. 9515d1b1b3fSAneesh Kumar K.V */ 9525d1b1b3fSAneesh Kumar K.V ext4_lock_group(sb, grp); 9535d1b1b3fSAneesh Kumar K.V return; 9545d1b1b3fSAneesh Kumar K.V } 9555d1b1b3fSAneesh Kumar K.V 956db79e6d1SWang Shilong void ext4_mark_group_bitmap_corrupted(struct super_block *sb, 957db79e6d1SWang Shilong ext4_group_t group, 958db79e6d1SWang Shilong unsigned int flags) 959db79e6d1SWang Shilong { 960db79e6d1SWang Shilong struct ext4_sb_info *sbi = EXT4_SB(sb); 961db79e6d1SWang Shilong struct ext4_group_info *grp = ext4_get_group_info(sb, group); 962db79e6d1SWang Shilong struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 9639af0b3d1SWang Shilong int ret; 964db79e6d1SWang Shilong 9659af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) { 9669af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT, 9679af0b3d1SWang Shilong &grp->bb_state); 9689af0b3d1SWang Shilong if (!ret) 969db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeclusters_counter, 970db79e6d1SWang Shilong grp->bb_free); 971db79e6d1SWang Shilong } 972db79e6d1SWang Shilong 9739af0b3d1SWang Shilong if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) { 9749af0b3d1SWang Shilong ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT, 9759af0b3d1SWang Shilong &grp->bb_state); 9769af0b3d1SWang Shilong if (!ret && gdp) { 977db79e6d1SWang Shilong int count; 978db79e6d1SWang Shilong 979db79e6d1SWang Shilong count = ext4_free_inodes_count(sb, gdp); 980db79e6d1SWang Shilong percpu_counter_sub(&sbi->s_freeinodes_counter, 981db79e6d1SWang Shilong count); 982db79e6d1SWang Shilong } 983db79e6d1SWang Shilong } 984db79e6d1SWang Shilong } 985db79e6d1SWang Shilong 986617ba13bSMingming Cao void ext4_update_dynamic_rev(struct super_block *sb) 987ac27a0ecSDave Kleikamp { 988617ba13bSMingming Cao struct ext4_super_block *es = EXT4_SB(sb)->s_es; 989ac27a0ecSDave Kleikamp 990617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 991ac27a0ecSDave Kleikamp return; 992ac27a0ecSDave Kleikamp 99312062dddSEric Sandeen ext4_warning(sb, 994ac27a0ecSDave Kleikamp "updating to rev %d because of new feature flag, " 995ac27a0ecSDave Kleikamp "running e2fsck is recommended", 996617ba13bSMingming Cao EXT4_DYNAMIC_REV); 997ac27a0ecSDave Kleikamp 998617ba13bSMingming Cao es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO); 999617ba13bSMingming Cao es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE); 1000617ba13bSMingming Cao es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV); 1001ac27a0ecSDave Kleikamp /* leave es->s_feature_*compat flags alone */ 1002ac27a0ecSDave Kleikamp /* es->s_uuid will be set by e2fsck if empty */ 1003ac27a0ecSDave Kleikamp 1004ac27a0ecSDave Kleikamp /* 1005ac27a0ecSDave Kleikamp * The rest of the superblock fields should be zero, and if not it 1006ac27a0ecSDave Kleikamp * means they are likely already in use, so leave them alone. We 1007ac27a0ecSDave Kleikamp * can leave it up to e2fsck to clean up any inconsistencies there. 1008ac27a0ecSDave Kleikamp */ 1009ac27a0ecSDave Kleikamp } 1010ac27a0ecSDave Kleikamp 1011ac27a0ecSDave Kleikamp /* 1012ac27a0ecSDave Kleikamp * Open the external journal device 1013ac27a0ecSDave Kleikamp */ 1014b31e1552SEric Sandeen static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb) 1015ac27a0ecSDave Kleikamp { 1016ac27a0ecSDave Kleikamp struct block_device *bdev; 1017ac27a0ecSDave Kleikamp 1018d4d77629STejun Heo bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb); 1019ac27a0ecSDave Kleikamp if (IS_ERR(bdev)) 1020ac27a0ecSDave Kleikamp goto fail; 1021ac27a0ecSDave Kleikamp return bdev; 1022ac27a0ecSDave Kleikamp 1023ac27a0ecSDave Kleikamp fail: 1024ea3edd4dSChristoph Hellwig ext4_msg(sb, KERN_ERR, 1025ea3edd4dSChristoph Hellwig "failed to open journal device unknown-block(%u,%u) %ld", 1026ea3edd4dSChristoph Hellwig MAJOR(dev), MINOR(dev), PTR_ERR(bdev)); 1027ac27a0ecSDave Kleikamp return NULL; 1028ac27a0ecSDave Kleikamp } 1029ac27a0ecSDave Kleikamp 1030ac27a0ecSDave Kleikamp /* 1031ac27a0ecSDave Kleikamp * Release the journal device 1032ac27a0ecSDave Kleikamp */ 10334385bab1SAl Viro static void ext4_blkdev_put(struct block_device *bdev) 1034ac27a0ecSDave Kleikamp { 10354385bab1SAl Viro blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL); 1036ac27a0ecSDave Kleikamp } 1037ac27a0ecSDave Kleikamp 10384385bab1SAl Viro static void ext4_blkdev_remove(struct ext4_sb_info *sbi) 1039ac27a0ecSDave Kleikamp { 1040ac27a0ecSDave Kleikamp struct block_device *bdev; 1041ee7ed3aaSChunguang Xu bdev = sbi->s_journal_bdev; 1042ac27a0ecSDave Kleikamp if (bdev) { 10434385bab1SAl Viro ext4_blkdev_put(bdev); 1044ee7ed3aaSChunguang Xu sbi->s_journal_bdev = NULL; 1045ac27a0ecSDave Kleikamp } 1046ac27a0ecSDave Kleikamp } 1047ac27a0ecSDave Kleikamp 1048ac27a0ecSDave Kleikamp static inline struct inode *orphan_list_entry(struct list_head *l) 1049ac27a0ecSDave Kleikamp { 1050617ba13bSMingming Cao return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode; 1051ac27a0ecSDave Kleikamp } 1052ac27a0ecSDave Kleikamp 1053617ba13bSMingming Cao static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi) 1054ac27a0ecSDave Kleikamp { 1055ac27a0ecSDave Kleikamp struct list_head *l; 1056ac27a0ecSDave Kleikamp 1057b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "sb orphan head is %d", 1058ac27a0ecSDave Kleikamp le32_to_cpu(sbi->s_es->s_last_orphan)); 1059ac27a0ecSDave Kleikamp 1060ac27a0ecSDave Kleikamp printk(KERN_ERR "sb_info orphan list:\n"); 1061ac27a0ecSDave Kleikamp list_for_each(l, &sbi->s_orphan) { 1062ac27a0ecSDave Kleikamp struct inode *inode = orphan_list_entry(l); 1063ac27a0ecSDave Kleikamp printk(KERN_ERR " " 1064ac27a0ecSDave Kleikamp "inode %s:%lu at %p: mode %o, nlink %d, next %d\n", 1065ac27a0ecSDave Kleikamp inode->i_sb->s_id, inode->i_ino, inode, 1066ac27a0ecSDave Kleikamp inode->i_mode, inode->i_nlink, 1067ac27a0ecSDave Kleikamp NEXT_ORPHAN(inode)); 1068ac27a0ecSDave Kleikamp } 1069ac27a0ecSDave Kleikamp } 1070ac27a0ecSDave Kleikamp 1071957153fcSJan Kara #ifdef CONFIG_QUOTA 1072957153fcSJan Kara static int ext4_quota_off(struct super_block *sb, int type); 1073957153fcSJan Kara 1074957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 1075957153fcSJan Kara { 1076957153fcSJan Kara int type; 1077957153fcSJan Kara 1078957153fcSJan Kara /* Use our quota_off function to clear inode flags etc. */ 1079957153fcSJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) 1080957153fcSJan Kara ext4_quota_off(sb, type); 1081957153fcSJan Kara } 108233458eabSTheodore Ts'o 108333458eabSTheodore Ts'o /* 108433458eabSTheodore Ts'o * This is a helper function which is used in the mount/remount 108533458eabSTheodore Ts'o * codepaths (which holds s_umount) to fetch the quota file name. 108633458eabSTheodore Ts'o */ 108733458eabSTheodore Ts'o static inline char *get_qf_name(struct super_block *sb, 108833458eabSTheodore Ts'o struct ext4_sb_info *sbi, 108933458eabSTheodore Ts'o int type) 109033458eabSTheodore Ts'o { 109133458eabSTheodore Ts'o return rcu_dereference_protected(sbi->s_qf_names[type], 109233458eabSTheodore Ts'o lockdep_is_held(&sb->s_umount)); 109333458eabSTheodore Ts'o } 1094957153fcSJan Kara #else 1095957153fcSJan Kara static inline void ext4_quota_off_umount(struct super_block *sb) 1096957153fcSJan Kara { 1097957153fcSJan Kara } 1098957153fcSJan Kara #endif 1099957153fcSJan Kara 1100617ba13bSMingming Cao static void ext4_put_super(struct super_block *sb) 1101ac27a0ecSDave Kleikamp { 1102617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 1103617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 11041d0c3924STheodore Ts'o struct buffer_head **group_desc; 11057c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 110697abd7d4STheodore Ts'o int aborted = 0; 1107ef2cabf7SHidehiro Kawai int i, err; 1108ac27a0ecSDave Kleikamp 1109857ac889SLukas Czerner ext4_unregister_li_request(sb); 1110957153fcSJan Kara ext4_quota_off_umount(sb); 1111e0ccfd95SChristoph Hellwig 11122e8fa54eSJan Kara destroy_workqueue(sbi->rsv_conversion_wq); 11134c0425ffSMingming Cao 11145e47868fSRitesh Harjani /* 11155e47868fSRitesh Harjani * Unregister sysfs before destroying jbd2 journal. 11165e47868fSRitesh Harjani * Since we could still access attr_journal_task attribute via sysfs 11175e47868fSRitesh Harjani * path which could have sbi->s_journal->j_task as NULL 11185e47868fSRitesh Harjani */ 11195e47868fSRitesh Harjani ext4_unregister_sysfs(sb); 11205e47868fSRitesh Harjani 11210390131bSFrank Mayhar if (sbi->s_journal) { 112297abd7d4STheodore Ts'o aborted = is_journal_aborted(sbi->s_journal); 1123ef2cabf7SHidehiro Kawai err = jbd2_journal_destroy(sbi->s_journal); 112447b4a50bSJan Kara sbi->s_journal = NULL; 1125878520acSTheodore Ts'o if ((err < 0) && !aborted) { 112654d3adbcSTheodore Ts'o ext4_abort(sb, -err, "Couldn't clean up the journal"); 11270390131bSFrank Mayhar } 1128878520acSTheodore Ts'o } 1129d4edac31SJosef Bacik 1130d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 11319105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 1132d4edac31SJosef Bacik ext4_release_system_zone(sb); 1133d4edac31SJosef Bacik ext4_mb_release(sb); 1134d4edac31SJosef Bacik ext4_ext_release(sb); 1135d4edac31SJosef Bacik 1136bc98a42cSDavid Howells if (!sb_rdonly(sb) && !aborted) { 1137e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 1138ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 1139ac27a0ecSDave Kleikamp } 1140bc98a42cSDavid Howells if (!sb_rdonly(sb)) 1141a8e25a83SArtem Bityutskiy ext4_commit_super(sb, 1); 1142a8e25a83SArtem Bityutskiy 11431d0c3924STheodore Ts'o rcu_read_lock(); 11441d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 1145ac27a0ecSDave Kleikamp for (i = 0; i < sbi->s_gdb_count; i++) 11461d0c3924STheodore Ts'o brelse(group_desc[i]); 11471d0c3924STheodore Ts'o kvfree(group_desc); 11487c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 11497c990728SSuraj Jitindar Singh if (flex_groups) { 11507c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 11517c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 11527c990728SSuraj Jitindar Singh kvfree(flex_groups); 11537c990728SSuraj Jitindar Singh } 11541d0c3924STheodore Ts'o rcu_read_unlock(); 115557042651STheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 1156ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_freeinodes_counter); 1157ac27a0ecSDave Kleikamp percpu_counter_destroy(&sbi->s_dirs_counter); 115857042651STheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 1159bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 1160ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1161a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 116233458eabSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 1163ac27a0ecSDave Kleikamp #endif 1164ac27a0ecSDave Kleikamp 1165ac27a0ecSDave Kleikamp /* Debugging code just in case the in-memory inode orphan list 1166ac27a0ecSDave Kleikamp * isn't empty. The on-disk one can be non-empty if we've 1167ac27a0ecSDave Kleikamp * detected an error and taken the fs readonly, but the 1168ac27a0ecSDave Kleikamp * in-memory list had better be clean by this point. */ 1169ac27a0ecSDave Kleikamp if (!list_empty(&sbi->s_orphan)) 1170ac27a0ecSDave Kleikamp dump_orphan_list(sb, sbi); 1171ac27a0ecSDave Kleikamp J_ASSERT(list_empty(&sbi->s_orphan)); 1172ac27a0ecSDave Kleikamp 117389d96a6fSTheodore Ts'o sync_blockdev(sb->s_bdev); 1174f98393a6SPeter Zijlstra invalidate_bdev(sb->s_bdev); 1175ee7ed3aaSChunguang Xu if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) { 1176ac27a0ecSDave Kleikamp /* 1177ac27a0ecSDave Kleikamp * Invalidate the journal device's buffers. We don't want them 1178ac27a0ecSDave Kleikamp * floating about in memory - the physical journal device may 1179ac27a0ecSDave Kleikamp * hotswapped, and it breaks the `ro-after' testing code. 1180ac27a0ecSDave Kleikamp */ 1181ee7ed3aaSChunguang Xu sync_blockdev(sbi->s_journal_bdev); 1182ee7ed3aaSChunguang Xu invalidate_bdev(sbi->s_journal_bdev); 1183617ba13bSMingming Cao ext4_blkdev_remove(sbi); 1184ac27a0ecSDave Kleikamp } 118550c15df6SChengguang Xu 1186dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 1187dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 118850c15df6SChengguang Xu 118947387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 119047387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 119150c15df6SChengguang Xu 1192c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 1193c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 11949060dd2cSEric Sandeen brelse(sbi->s_sbh); 1195ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 11963197ebdbSTheodore Ts'o /* 11973197ebdbSTheodore Ts'o * Now that we are completely done shutting down the 11983197ebdbSTheodore Ts'o * superblock, we need to actually destroy the kobject. 11993197ebdbSTheodore Ts'o */ 12003197ebdbSTheodore Ts'o kobject_put(&sbi->s_kobj); 12013197ebdbSTheodore Ts'o wait_for_completion(&sbi->s_kobj_unregister); 12020441984aSDarrick J. Wong if (sbi->s_chksum_driver) 12030441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 1204705895b6SPekka Enberg kfree(sbi->s_blockgroup_lock); 12055e405595SDan Williams fs_put_dax(sbi->s_daxdev); 1206ed318a6cSEric Biggers fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx); 1207c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1208c83ad55eSGabriel Krisman Bertazi utf8_unload(sbi->s_encoding); 1209c83ad55eSGabriel Krisman Bertazi #endif 1210ac27a0ecSDave Kleikamp kfree(sbi); 1211ac27a0ecSDave Kleikamp } 1212ac27a0ecSDave Kleikamp 1213e18b890bSChristoph Lameter static struct kmem_cache *ext4_inode_cachep; 1214ac27a0ecSDave Kleikamp 1215ac27a0ecSDave Kleikamp /* 1216ac27a0ecSDave Kleikamp * Called inside transaction, so use GFP_NOFS 1217ac27a0ecSDave Kleikamp */ 1218617ba13bSMingming Cao static struct inode *ext4_alloc_inode(struct super_block *sb) 1219ac27a0ecSDave Kleikamp { 1220617ba13bSMingming Cao struct ext4_inode_info *ei; 1221ac27a0ecSDave Kleikamp 1222e6b4f8daSChristoph Lameter ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS); 1223ac27a0ecSDave Kleikamp if (!ei) 1224ac27a0ecSDave Kleikamp return NULL; 12250b8e58a1SAndreas Dilger 1226ee73f9a5SJeff Layton inode_set_iversion(&ei->vfs_inode, 1); 1227202ee5dfSTheodore Ts'o spin_lock_init(&ei->i_raw_lock); 1228c9de560dSAlex Tomas INIT_LIST_HEAD(&ei->i_prealloc_list); 122927bc446eSbrookxu atomic_set(&ei->i_prealloc_active, 0); 1230c9de560dSAlex Tomas spin_lock_init(&ei->i_prealloc_lock); 12319a26b661SZheng Liu ext4_es_init_tree(&ei->i_es_tree); 12329a26b661SZheng Liu rwlock_init(&ei->i_es_lock); 1233edaa53caSZheng Liu INIT_LIST_HEAD(&ei->i_es_list); 1234eb68d0e2SZheng Liu ei->i_es_all_nr = 0; 1235edaa53caSZheng Liu ei->i_es_shk_nr = 0; 1236dd475925SJan Kara ei->i_es_shrink_lblk = 0; 1237d2a17637SMingming Cao ei->i_reserved_data_blocks = 0; 1238d2a17637SMingming Cao spin_lock_init(&(ei->i_block_reservation_lock)); 12391dc0aa46SEric Whitney ext4_init_pending_tree(&ei->i_pending_tree); 1240a9e7f447SDmitry Monakhov #ifdef CONFIG_QUOTA 1241a9e7f447SDmitry Monakhov ei->i_reserved_quota = 0; 124296c7e0d9SJan Kara memset(&ei->i_dquot, 0, sizeof(ei->i_dquot)); 1243a9e7f447SDmitry Monakhov #endif 12448aefcd55STheodore Ts'o ei->jinode = NULL; 12452e8fa54eSJan Kara INIT_LIST_HEAD(&ei->i_rsv_conversion_list); 1246744692dcSJiaying Zhang spin_lock_init(&ei->i_completed_io_lock); 1247b436b9beSJan Kara ei->i_sync_tid = 0; 1248b436b9beSJan Kara ei->i_datasync_tid = 0; 1249e27f41e1SDmitry Monakhov atomic_set(&ei->i_unwritten, 0); 12502e8fa54eSJan Kara INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work); 1251ac27a0ecSDave Kleikamp return &ei->vfs_inode; 1252ac27a0ecSDave Kleikamp } 1253ac27a0ecSDave Kleikamp 12547ff9c073STheodore Ts'o static int ext4_drop_inode(struct inode *inode) 12557ff9c073STheodore Ts'o { 12567ff9c073STheodore Ts'o int drop = generic_drop_inode(inode); 12577ff9c073STheodore Ts'o 125829b3692eSEric Biggers if (!drop) 125929b3692eSEric Biggers drop = fscrypt_drop_inode(inode); 126029b3692eSEric Biggers 12617ff9c073STheodore Ts'o trace_ext4_drop_inode(inode, drop); 12627ff9c073STheodore Ts'o return drop; 12637ff9c073STheodore Ts'o } 12647ff9c073STheodore Ts'o 126594053139SAl Viro static void ext4_free_in_core_inode(struct inode *inode) 1266fa0d7e3dSNick Piggin { 12672c58d548SEric Biggers fscrypt_free_inode(inode); 1268fa0d7e3dSNick Piggin kmem_cache_free(ext4_inode_cachep, EXT4_I(inode)); 1269fa0d7e3dSNick Piggin } 1270fa0d7e3dSNick Piggin 1271617ba13bSMingming Cao static void ext4_destroy_inode(struct inode *inode) 1272ac27a0ecSDave Kleikamp { 12739f7dd93dSVasily Averin if (!list_empty(&(EXT4_I(inode)->i_orphan))) { 1274b31e1552SEric Sandeen ext4_msg(inode->i_sb, KERN_ERR, 1275b31e1552SEric Sandeen "Inode %lu (%p): orphan list check failed!", 1276b31e1552SEric Sandeen inode->i_ino, EXT4_I(inode)); 12779f7dd93dSVasily Averin print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4, 12789f7dd93dSVasily Averin EXT4_I(inode), sizeof(struct ext4_inode_info), 12799f7dd93dSVasily Averin true); 12809f7dd93dSVasily Averin dump_stack(); 12819f7dd93dSVasily Averin } 1282ac27a0ecSDave Kleikamp } 1283ac27a0ecSDave Kleikamp 128451cc5068SAlexey Dobriyan static void init_once(void *foo) 1285ac27a0ecSDave Kleikamp { 1286617ba13bSMingming Cao struct ext4_inode_info *ei = (struct ext4_inode_info *) foo; 1287ac27a0ecSDave Kleikamp 1288ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&ei->i_orphan); 1289ac27a0ecSDave Kleikamp init_rwsem(&ei->xattr_sem); 12900e855ac8SAneesh Kumar K.V init_rwsem(&ei->i_data_sem); 1291ea3d7209SJan Kara init_rwsem(&ei->i_mmap_sem); 1292ac27a0ecSDave Kleikamp inode_init_once(&ei->vfs_inode); 1293ac27a0ecSDave Kleikamp } 1294ac27a0ecSDave Kleikamp 1295e67bc2b3SFabian Frederick static int __init init_inodecache(void) 1296ac27a0ecSDave Kleikamp { 1297f8dd7c70SDavid Windsor ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache", 1298f8dd7c70SDavid Windsor sizeof(struct ext4_inode_info), 0, 1299f8dd7c70SDavid Windsor (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| 1300f8dd7c70SDavid Windsor SLAB_ACCOUNT), 1301f8dd7c70SDavid Windsor offsetof(struct ext4_inode_info, i_data), 1302f8dd7c70SDavid Windsor sizeof_field(struct ext4_inode_info, i_data), 130320c2df83SPaul Mundt init_once); 1304617ba13bSMingming Cao if (ext4_inode_cachep == NULL) 1305ac27a0ecSDave Kleikamp return -ENOMEM; 1306ac27a0ecSDave Kleikamp return 0; 1307ac27a0ecSDave Kleikamp } 1308ac27a0ecSDave Kleikamp 1309ac27a0ecSDave Kleikamp static void destroy_inodecache(void) 1310ac27a0ecSDave Kleikamp { 13118c0a8537SKirill A. Shutemov /* 13128c0a8537SKirill A. Shutemov * Make sure all delayed rcu free inodes are flushed before we 13138c0a8537SKirill A. Shutemov * destroy cache. 13148c0a8537SKirill A. Shutemov */ 13158c0a8537SKirill A. Shutemov rcu_barrier(); 1316617ba13bSMingming Cao kmem_cache_destroy(ext4_inode_cachep); 1317ac27a0ecSDave Kleikamp } 1318ac27a0ecSDave Kleikamp 13190930fcc1SAl Viro void ext4_clear_inode(struct inode *inode) 1320ac27a0ecSDave Kleikamp { 13210930fcc1SAl Viro invalidate_inode_buffers(inode); 1322dbd5768fSJan Kara clear_inode(inode); 132327bc446eSbrookxu ext4_discard_preallocations(inode, 0); 132451865fdaSZheng Liu ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS); 1325f4c2d372SJan Kara dquot_drop(inode); 13268aefcd55STheodore Ts'o if (EXT4_I(inode)->jinode) { 13278aefcd55STheodore Ts'o jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode), 13288aefcd55STheodore Ts'o EXT4_I(inode)->jinode); 13298aefcd55STheodore Ts'o jbd2_free_inode(EXT4_I(inode)->jinode); 13308aefcd55STheodore Ts'o EXT4_I(inode)->jinode = NULL; 13318aefcd55STheodore Ts'o } 13323d204e24SEric Biggers fscrypt_put_encryption_info(inode); 1333c93d8f88SEric Biggers fsverity_cleanup_inode(inode); 1334ac27a0ecSDave Kleikamp } 1335ac27a0ecSDave Kleikamp 13361b961ac0SChristoph Hellwig static struct inode *ext4_nfs_get_inode(struct super_block *sb, 13371b961ac0SChristoph Hellwig u64 ino, u32 generation) 1338ac27a0ecSDave Kleikamp { 1339ac27a0ecSDave Kleikamp struct inode *inode; 1340ac27a0ecSDave Kleikamp 13418a363970STheodore Ts'o /* 1342ac27a0ecSDave Kleikamp * Currently we don't know the generation for parent directory, so 1343ac27a0ecSDave Kleikamp * a generation of 0 means "accept any" 1344ac27a0ecSDave Kleikamp */ 13458a363970STheodore Ts'o inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE); 13461d1fe1eeSDavid Howells if (IS_ERR(inode)) 13471d1fe1eeSDavid Howells return ERR_CAST(inode); 13481d1fe1eeSDavid Howells if (generation && inode->i_generation != generation) { 1349ac27a0ecSDave Kleikamp iput(inode); 1350ac27a0ecSDave Kleikamp return ERR_PTR(-ESTALE); 1351ac27a0ecSDave Kleikamp } 13521b961ac0SChristoph Hellwig 13531b961ac0SChristoph Hellwig return inode; 1354ac27a0ecSDave Kleikamp } 13551b961ac0SChristoph Hellwig 13561b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid, 13571b961ac0SChristoph Hellwig int fh_len, int fh_type) 13581b961ac0SChristoph Hellwig { 13591b961ac0SChristoph Hellwig return generic_fh_to_dentry(sb, fid, fh_len, fh_type, 13601b961ac0SChristoph Hellwig ext4_nfs_get_inode); 13611b961ac0SChristoph Hellwig } 13621b961ac0SChristoph Hellwig 13631b961ac0SChristoph Hellwig static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid, 13641b961ac0SChristoph Hellwig int fh_len, int fh_type) 13651b961ac0SChristoph Hellwig { 13661b961ac0SChristoph Hellwig return generic_fh_to_parent(sb, fid, fh_len, fh_type, 13671b961ac0SChristoph Hellwig ext4_nfs_get_inode); 1368ac27a0ecSDave Kleikamp } 1369ac27a0ecSDave Kleikamp 1370fde87268STheodore Ts'o static int ext4_nfs_commit_metadata(struct inode *inode) 1371fde87268STheodore Ts'o { 1372fde87268STheodore Ts'o struct writeback_control wbc = { 1373fde87268STheodore Ts'o .sync_mode = WB_SYNC_ALL 1374fde87268STheodore Ts'o }; 1375fde87268STheodore Ts'o 1376fde87268STheodore Ts'o trace_ext4_nfs_commit_metadata(inode); 1377fde87268STheodore Ts'o return ext4_write_inode(inode, &wbc); 1378fde87268STheodore Ts'o } 1379fde87268STheodore Ts'o 1380c39a7f84SToshiyuki Okajima /* 1381c39a7f84SToshiyuki Okajima * Try to release metadata pages (indirect blocks, directories) which are 1382c39a7f84SToshiyuki Okajima * mapped via the block device. Since these pages could have journal heads 1383c39a7f84SToshiyuki Okajima * which would prevent try_to_free_buffers() from freeing them, we must use 1384c39a7f84SToshiyuki Okajima * jbd2 layer's try_to_free_buffers() function to release them. 1385c39a7f84SToshiyuki Okajima */ 13860b8e58a1SAndreas Dilger static int bdev_try_to_free_page(struct super_block *sb, struct page *page, 13870b8e58a1SAndreas Dilger gfp_t wait) 1388c39a7f84SToshiyuki Okajima { 1389c39a7f84SToshiyuki Okajima journal_t *journal = EXT4_SB(sb)->s_journal; 1390c39a7f84SToshiyuki Okajima 1391c39a7f84SToshiyuki Okajima WARN_ON(PageChecked(page)); 1392c39a7f84SToshiyuki Okajima if (!page_has_buffers(page)) 1393c39a7f84SToshiyuki Okajima return 0; 1394c39a7f84SToshiyuki Okajima if (journal) 1395529a781eSzhangyi (F) return jbd2_journal_try_to_free_buffers(journal, page); 1396529a781eSzhangyi (F) 1397c39a7f84SToshiyuki Okajima return try_to_free_buffers(page); 1398c39a7f84SToshiyuki Okajima } 1399c39a7f84SToshiyuki Okajima 1400643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 1401a7550b30SJaegeuk Kim static int ext4_get_context(struct inode *inode, void *ctx, size_t len) 1402a7550b30SJaegeuk Kim { 1403a7550b30SJaegeuk Kim return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION, 1404a7550b30SJaegeuk Kim EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len); 1405a7550b30SJaegeuk Kim } 1406a7550b30SJaegeuk Kim 1407a7550b30SJaegeuk Kim static int ext4_set_context(struct inode *inode, const void *ctx, size_t len, 1408a7550b30SJaegeuk Kim void *fs_data) 1409a7550b30SJaegeuk Kim { 14102f8f5e76SEric Biggers handle_t *handle = fs_data; 1411c1a5d5f6STahsin Erdogan int res, res2, credits, retries = 0; 1412a7550b30SJaegeuk Kim 14139ce0151aSEric Biggers /* 14149ce0151aSEric Biggers * Encrypting the root directory is not allowed because e2fsck expects 14159ce0151aSEric Biggers * lost+found to exist and be unencrypted, and encrypting the root 14169ce0151aSEric Biggers * directory would imply encrypting the lost+found directory as well as 14179ce0151aSEric Biggers * the filename "lost+found" itself. 14189ce0151aSEric Biggers */ 14199ce0151aSEric Biggers if (inode->i_ino == EXT4_ROOT_INO) 14209ce0151aSEric Biggers return -EPERM; 1421a7550b30SJaegeuk Kim 14227d3e06a8SRoss Zwisler if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode))) 14237d3e06a8SRoss Zwisler return -EINVAL; 14247d3e06a8SRoss Zwisler 1425b383a73fSIra Weiny if (ext4_test_inode_flag(inode, EXT4_INODE_DAX)) 1426b383a73fSIra Weiny return -EOPNOTSUPP; 1427b383a73fSIra Weiny 142894840e3cSEric Biggers res = ext4_convert_inline_data(inode); 142994840e3cSEric Biggers if (res) 143094840e3cSEric Biggers return res; 143194840e3cSEric Biggers 14322f8f5e76SEric Biggers /* 14332f8f5e76SEric Biggers * If a journal handle was specified, then the encryption context is 14342f8f5e76SEric Biggers * being set on a new inode via inheritance and is part of a larger 14352f8f5e76SEric Biggers * transaction to create the inode. Otherwise the encryption context is 14362f8f5e76SEric Biggers * being set on an existing inode in its own transaction. Only in the 14372f8f5e76SEric Biggers * latter case should the "retry on ENOSPC" logic be used. 14382f8f5e76SEric Biggers */ 14392f8f5e76SEric Biggers 14402f8f5e76SEric Biggers if (handle) { 14412f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, 14422f8f5e76SEric Biggers EXT4_XATTR_INDEX_ENCRYPTION, 14432f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 14442f8f5e76SEric Biggers ctx, len, 0); 1445a7550b30SJaegeuk Kim if (!res) { 1446a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 1447a7550b30SJaegeuk Kim ext4_clear_inode_state(inode, 1448a7550b30SJaegeuk Kim EXT4_STATE_MAY_INLINE_DATA); 1449a3caa24bSJan Kara /* 14502ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 14512ee6a576SEric Biggers * S_DAX may be disabled 1452a3caa24bSJan Kara */ 1453043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1454a7550b30SJaegeuk Kim } 1455a7550b30SJaegeuk Kim return res; 1456a7550b30SJaegeuk Kim } 1457a7550b30SJaegeuk Kim 1458b8cb5a54STahsin Erdogan res = dquot_initialize(inode); 1459b8cb5a54STahsin Erdogan if (res) 1460b8cb5a54STahsin Erdogan return res; 14612f8f5e76SEric Biggers retry: 1462af65207cSTahsin Erdogan res = ext4_xattr_set_credits(inode, len, false /* is_create */, 1463af65207cSTahsin Erdogan &credits); 1464dec214d0STahsin Erdogan if (res) 1465dec214d0STahsin Erdogan return res; 1466dec214d0STahsin Erdogan 1467c1a5d5f6STahsin Erdogan handle = ext4_journal_start(inode, EXT4_HT_MISC, credits); 1468a7550b30SJaegeuk Kim if (IS_ERR(handle)) 1469a7550b30SJaegeuk Kim return PTR_ERR(handle); 1470a7550b30SJaegeuk Kim 14712f8f5e76SEric Biggers res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION, 14722f8f5e76SEric Biggers EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, 14732f8f5e76SEric Biggers ctx, len, 0); 1474a7550b30SJaegeuk Kim if (!res) { 1475a7550b30SJaegeuk Kim ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT); 14762ee6a576SEric Biggers /* 14772ee6a576SEric Biggers * Update inode->i_flags - S_ENCRYPTED will be enabled, 14782ee6a576SEric Biggers * S_DAX may be disabled 14792ee6a576SEric Biggers */ 1480043546e4SIra Weiny ext4_set_inode_flags(inode, false); 1481a7550b30SJaegeuk Kim res = ext4_mark_inode_dirty(handle, inode); 1482a7550b30SJaegeuk Kim if (res) 1483a7550b30SJaegeuk Kim EXT4_ERROR_INODE(inode, "Failed to mark inode dirty"); 1484a7550b30SJaegeuk Kim } 1485a7550b30SJaegeuk Kim res2 = ext4_journal_stop(handle); 14862f8f5e76SEric Biggers 14872f8f5e76SEric Biggers if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 14882f8f5e76SEric Biggers goto retry; 1489a7550b30SJaegeuk Kim if (!res) 1490a7550b30SJaegeuk Kim res = res2; 1491a7550b30SJaegeuk Kim return res; 1492a7550b30SJaegeuk Kim } 1493a7550b30SJaegeuk Kim 1494ed318a6cSEric Biggers static const union fscrypt_context * 1495ed318a6cSEric Biggers ext4_get_dummy_context(struct super_block *sb) 1496a7550b30SJaegeuk Kim { 1497ed318a6cSEric Biggers return EXT4_SB(sb)->s_dummy_enc_ctx.ctx; 1498a7550b30SJaegeuk Kim } 1499a7550b30SJaegeuk Kim 1500b925acb8SEric Biggers static bool ext4_has_stable_inodes(struct super_block *sb) 1501b925acb8SEric Biggers { 1502b925acb8SEric Biggers return ext4_has_feature_stable_inodes(sb); 1503b925acb8SEric Biggers } 1504b925acb8SEric Biggers 1505b925acb8SEric Biggers static void ext4_get_ino_and_lblk_bits(struct super_block *sb, 1506b925acb8SEric Biggers int *ino_bits_ret, int *lblk_bits_ret) 1507b925acb8SEric Biggers { 1508b925acb8SEric Biggers *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count); 1509b925acb8SEric Biggers *lblk_bits_ret = 8 * sizeof(ext4_lblk_t); 1510b925acb8SEric Biggers } 1511b925acb8SEric Biggers 15126f69f0edSEric Biggers static const struct fscrypt_operations ext4_cryptops = { 1513a5d431efSEric Biggers .key_prefix = "ext4:", 1514a7550b30SJaegeuk Kim .get_context = ext4_get_context, 1515a7550b30SJaegeuk Kim .set_context = ext4_set_context, 1516ed318a6cSEric Biggers .get_dummy_context = ext4_get_dummy_context, 1517a7550b30SJaegeuk Kim .empty_dir = ext4_empty_dir, 1518e12ee683SEric Biggers .max_namelen = EXT4_NAME_LEN, 1519b925acb8SEric Biggers .has_stable_inodes = ext4_has_stable_inodes, 1520b925acb8SEric Biggers .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits, 1521a7550b30SJaegeuk Kim }; 1522a7550b30SJaegeuk Kim #endif 1523a7550b30SJaegeuk Kim 1524ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1525d6006186SEric Biggers static const char * const quotatypes[] = INITQFNAMES; 1526689c958cSLi Xi #define QTYPE2NAME(t) (quotatypes[t]) 1527ac27a0ecSDave Kleikamp 1528617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot); 1529617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot); 1530617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot); 1531617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot); 1532617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type); 15336f28e087SJan Kara static int ext4_quota_on(struct super_block *sb, int type, int format_id, 15348c54ca9cSAl Viro const struct path *path); 1535617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type); 1536617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1537ac27a0ecSDave Kleikamp size_t len, loff_t off); 1538617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 1539ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off); 15407c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 15417c319d32SAditya Kali unsigned int flags); 15427c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb); 1543ac27a0ecSDave Kleikamp 154496c7e0d9SJan Kara static struct dquot **ext4_get_dquots(struct inode *inode) 154596c7e0d9SJan Kara { 154696c7e0d9SJan Kara return EXT4_I(inode)->i_dquot; 154796c7e0d9SJan Kara } 154896c7e0d9SJan Kara 154961e225dcSAlexey Dobriyan static const struct dquot_operations ext4_quota_operations = { 155060e58e0fSMingming Cao .get_reserved_space = ext4_get_reserved_space, 1551617ba13bSMingming Cao .write_dquot = ext4_write_dquot, 1552617ba13bSMingming Cao .acquire_dquot = ext4_acquire_dquot, 1553617ba13bSMingming Cao .release_dquot = ext4_release_dquot, 1554617ba13bSMingming Cao .mark_dirty = ext4_mark_dquot_dirty, 1555a5b5ee32SJan Kara .write_info = ext4_write_info, 1556a5b5ee32SJan Kara .alloc_dquot = dquot_alloc, 1557a5b5ee32SJan Kara .destroy_dquot = dquot_destroy, 1558040cb378SLi Xi .get_projid = ext4_get_projid, 15597a9ca53aSTahsin Erdogan .get_inode_usage = ext4_get_inode_usage, 1560ebc11f7bSChengguang Xu .get_next_id = dquot_get_next_id, 1561ac27a0ecSDave Kleikamp }; 1562ac27a0ecSDave Kleikamp 15630d54b217SAlexey Dobriyan static const struct quotactl_ops ext4_qctl_operations = { 1564617ba13bSMingming Cao .quota_on = ext4_quota_on, 1565ca0e05e4SDmitry Monakhov .quota_off = ext4_quota_off, 1566287a8095SChristoph Hellwig .quota_sync = dquot_quota_sync, 15670a240339SJan Kara .get_state = dquot_get_state, 1568287a8095SChristoph Hellwig .set_info = dquot_set_dqinfo, 1569287a8095SChristoph Hellwig .get_dqblk = dquot_get_dqblk, 15706332b9b5SEric Sandeen .set_dqblk = dquot_set_dqblk, 15716332b9b5SEric Sandeen .get_nextdqblk = dquot_get_next_dqblk, 1572ac27a0ecSDave Kleikamp }; 1573ac27a0ecSDave Kleikamp #endif 1574ac27a0ecSDave Kleikamp 1575ee9b6d61SJosef 'Jeff' Sipek static const struct super_operations ext4_sops = { 1576617ba13bSMingming Cao .alloc_inode = ext4_alloc_inode, 157794053139SAl Viro .free_inode = ext4_free_in_core_inode, 1578617ba13bSMingming Cao .destroy_inode = ext4_destroy_inode, 1579617ba13bSMingming Cao .write_inode = ext4_write_inode, 1580617ba13bSMingming Cao .dirty_inode = ext4_dirty_inode, 15817ff9c073STheodore Ts'o .drop_inode = ext4_drop_inode, 15820930fcc1SAl Viro .evict_inode = ext4_evict_inode, 1583617ba13bSMingming Cao .put_super = ext4_put_super, 1584617ba13bSMingming Cao .sync_fs = ext4_sync_fs, 1585c4be0c1dSTakashi Sato .freeze_fs = ext4_freeze, 1586c4be0c1dSTakashi Sato .unfreeze_fs = ext4_unfreeze, 1587617ba13bSMingming Cao .statfs = ext4_statfs, 1588617ba13bSMingming Cao .remount_fs = ext4_remount, 1589617ba13bSMingming Cao .show_options = ext4_show_options, 1590ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 1591617ba13bSMingming Cao .quota_read = ext4_quota_read, 1592617ba13bSMingming Cao .quota_write = ext4_quota_write, 159396c7e0d9SJan Kara .get_dquots = ext4_get_dquots, 1594ac27a0ecSDave Kleikamp #endif 1595c39a7f84SToshiyuki Okajima .bdev_try_to_free_page = bdev_try_to_free_page, 1596ac27a0ecSDave Kleikamp }; 1597ac27a0ecSDave Kleikamp 159839655164SChristoph Hellwig static const struct export_operations ext4_export_ops = { 15991b961ac0SChristoph Hellwig .fh_to_dentry = ext4_fh_to_dentry, 16001b961ac0SChristoph Hellwig .fh_to_parent = ext4_fh_to_parent, 1601617ba13bSMingming Cao .get_parent = ext4_get_parent, 1602fde87268STheodore Ts'o .commit_metadata = ext4_nfs_commit_metadata, 1603ac27a0ecSDave Kleikamp }; 1604ac27a0ecSDave Kleikamp 1605ac27a0ecSDave Kleikamp enum { 1606ac27a0ecSDave Kleikamp Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid, 1607ac27a0ecSDave Kleikamp Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro, 160872578c33STheodore Ts'o Opt_nouid32, Opt_debug, Opt_removed, 1609ac27a0ecSDave Kleikamp Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl, 161072578c33STheodore Ts'o Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, 1611ad4eec61SEric Sandeen Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev, 1612ad4eec61SEric Sandeen Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit, 1613ac27a0ecSDave Kleikamp Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback, 16146ddb2447STheodore Ts'o Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption, 16154f74d15fSEric Biggers Opt_inlinecrypt, 1616ac27a0ecSDave Kleikamp Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota, 16175a20bdfcSJan Kara Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota, 1618ee4a3fcdSTheodore Ts'o Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err, 16199cb20f94SIra Weiny Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, 16209cb20f94SIra Weiny Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never, 1621327eaf73STheodore Ts'o Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error, 1622327eaf73STheodore Ts'o Opt_nowarn_on_error, Opt_mblk_io_submit, 1623670e9875STheodore Ts'o Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize, 16241449032bSTheodore Ts'o Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity, 16255328e635SEric Sandeen Opt_inode_readahead_blks, Opt_journal_ioprio, 1626744692dcSJiaying Zhang Opt_dioread_nolock, Opt_dioread_lock, 1627fc6cb1cdSTheodore Ts'o Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable, 1628cdb7ee4cSTahsin Erdogan Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache, 16293d392b26STheodore Ts'o Opt_prefetch_block_bitmaps, 1630ac27a0ecSDave Kleikamp }; 1631ac27a0ecSDave Kleikamp 1632a447c093SSteven Whitehouse static const match_table_t tokens = { 1633ac27a0ecSDave Kleikamp {Opt_bsd_df, "bsddf"}, 1634ac27a0ecSDave Kleikamp {Opt_minix_df, "minixdf"}, 1635ac27a0ecSDave Kleikamp {Opt_grpid, "grpid"}, 1636ac27a0ecSDave Kleikamp {Opt_grpid, "bsdgroups"}, 1637ac27a0ecSDave Kleikamp {Opt_nogrpid, "nogrpid"}, 1638ac27a0ecSDave Kleikamp {Opt_nogrpid, "sysvgroups"}, 1639ac27a0ecSDave Kleikamp {Opt_resgid, "resgid=%u"}, 1640ac27a0ecSDave Kleikamp {Opt_resuid, "resuid=%u"}, 1641ac27a0ecSDave Kleikamp {Opt_sb, "sb=%u"}, 1642ac27a0ecSDave Kleikamp {Opt_err_cont, "errors=continue"}, 1643ac27a0ecSDave Kleikamp {Opt_err_panic, "errors=panic"}, 1644ac27a0ecSDave Kleikamp {Opt_err_ro, "errors=remount-ro"}, 1645ac27a0ecSDave Kleikamp {Opt_nouid32, "nouid32"}, 1646ac27a0ecSDave Kleikamp {Opt_debug, "debug"}, 164772578c33STheodore Ts'o {Opt_removed, "oldalloc"}, 164872578c33STheodore Ts'o {Opt_removed, "orlov"}, 1649ac27a0ecSDave Kleikamp {Opt_user_xattr, "user_xattr"}, 1650ac27a0ecSDave Kleikamp {Opt_nouser_xattr, "nouser_xattr"}, 1651ac27a0ecSDave Kleikamp {Opt_acl, "acl"}, 1652ac27a0ecSDave Kleikamp {Opt_noacl, "noacl"}, 1653e3bb52aeSEric Sandeen {Opt_noload, "norecovery"}, 16545a916be1STheodore Ts'o {Opt_noload, "noload"}, 165572578c33STheodore Ts'o {Opt_removed, "nobh"}, 165672578c33STheodore Ts'o {Opt_removed, "bh"}, 1657ac27a0ecSDave Kleikamp {Opt_commit, "commit=%u"}, 165830773840STheodore Ts'o {Opt_min_batch_time, "min_batch_time=%u"}, 165930773840STheodore Ts'o {Opt_max_batch_time, "max_batch_time=%u"}, 1660ac27a0ecSDave Kleikamp {Opt_journal_dev, "journal_dev=%u"}, 1661ad4eec61SEric Sandeen {Opt_journal_path, "journal_path=%s"}, 1662818d276cSGirish Shilamkar {Opt_journal_checksum, "journal_checksum"}, 1663c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, "nojournal_checksum"}, 1664818d276cSGirish Shilamkar {Opt_journal_async_commit, "journal_async_commit"}, 1665ac27a0ecSDave Kleikamp {Opt_abort, "abort"}, 1666ac27a0ecSDave Kleikamp {Opt_data_journal, "data=journal"}, 1667ac27a0ecSDave Kleikamp {Opt_data_ordered, "data=ordered"}, 1668ac27a0ecSDave Kleikamp {Opt_data_writeback, "data=writeback"}, 16695bf5683aSHidehiro Kawai {Opt_data_err_abort, "data_err=abort"}, 16705bf5683aSHidehiro Kawai {Opt_data_err_ignore, "data_err=ignore"}, 1671ac27a0ecSDave Kleikamp {Opt_offusrjquota, "usrjquota="}, 1672ac27a0ecSDave Kleikamp {Opt_usrjquota, "usrjquota=%s"}, 1673ac27a0ecSDave Kleikamp {Opt_offgrpjquota, "grpjquota="}, 1674ac27a0ecSDave Kleikamp {Opt_grpjquota, "grpjquota=%s"}, 1675ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsold, "jqfmt=vfsold"}, 1676ac27a0ecSDave Kleikamp {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"}, 16775a20bdfcSJan Kara {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"}, 1678ac27a0ecSDave Kleikamp {Opt_grpquota, "grpquota"}, 1679ac27a0ecSDave Kleikamp {Opt_noquota, "noquota"}, 1680ac27a0ecSDave Kleikamp {Opt_quota, "quota"}, 1681ac27a0ecSDave Kleikamp {Opt_usrquota, "usrquota"}, 168249da9392SJan Kara {Opt_prjquota, "prjquota"}, 1683ac27a0ecSDave Kleikamp {Opt_barrier, "barrier=%u"}, 168406705bffSTheodore Ts'o {Opt_barrier, "barrier"}, 168506705bffSTheodore Ts'o {Opt_nobarrier, "nobarrier"}, 168625ec56b5SJean Noel Cordenner {Opt_i_version, "i_version"}, 1687923ae0ffSRoss Zwisler {Opt_dax, "dax"}, 16889cb20f94SIra Weiny {Opt_dax_always, "dax=always"}, 16899cb20f94SIra Weiny {Opt_dax_inode, "dax=inode"}, 16909cb20f94SIra Weiny {Opt_dax_never, "dax=never"}, 1691c9de560dSAlex Tomas {Opt_stripe, "stripe=%u"}, 169264769240SAlex Tomas {Opt_delalloc, "delalloc"}, 1693327eaf73STheodore Ts'o {Opt_warn_on_error, "warn_on_error"}, 1694327eaf73STheodore Ts'o {Opt_nowarn_on_error, "nowarn_on_error"}, 1695a26f4992STheodore Ts'o {Opt_lazytime, "lazytime"}, 1696a26f4992STheodore Ts'o {Opt_nolazytime, "nolazytime"}, 1697670e9875STheodore Ts'o {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"}, 1698dd919b98SAneesh Kumar K.V {Opt_nodelalloc, "nodelalloc"}, 169936ade451SJan Kara {Opt_removed, "mblk_io_submit"}, 170036ade451SJan Kara {Opt_removed, "nomblk_io_submit"}, 17016fd058f7STheodore Ts'o {Opt_block_validity, "block_validity"}, 17026fd058f7STheodore Ts'o {Opt_noblock_validity, "noblock_validity"}, 1703240799cdSTheodore Ts'o {Opt_inode_readahead_blks, "inode_readahead_blks=%u"}, 1704b3881f74STheodore Ts'o {Opt_journal_ioprio, "journal_ioprio=%u"}, 1705afd4672dSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc=%u"}, 170606705bffSTheodore Ts'o {Opt_auto_da_alloc, "auto_da_alloc"}, 170706705bffSTheodore Ts'o {Opt_noauto_da_alloc, "noauto_da_alloc"}, 1708744692dcSJiaying Zhang {Opt_dioread_nolock, "dioread_nolock"}, 1709244adf64STheodore Ts'o {Opt_dioread_lock, "nodioread_nolock"}, 1710744692dcSJiaying Zhang {Opt_dioread_lock, "dioread_lock"}, 17115328e635SEric Sandeen {Opt_discard, "discard"}, 17125328e635SEric Sandeen {Opt_nodiscard, "nodiscard"}, 1713fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable=%u"}, 1714fc6cb1cdSTheodore Ts'o {Opt_init_itable, "init_itable"}, 1715fc6cb1cdSTheodore Ts'o {Opt_noinit_itable, "noinit_itable"}, 1716df981d03STheodore Ts'o {Opt_max_dir_size_kb, "max_dir_size_kb=%u"}, 1717ed318a6cSEric Biggers {Opt_test_dummy_encryption, "test_dummy_encryption=%s"}, 17186ddb2447STheodore Ts'o {Opt_test_dummy_encryption, "test_dummy_encryption"}, 17194f74d15fSEric Biggers {Opt_inlinecrypt, "inlinecrypt"}, 1720cdb7ee4cSTahsin Erdogan {Opt_nombcache, "nombcache"}, 1721cdb7ee4cSTahsin Erdogan {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */ 17223d392b26STheodore Ts'o {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"}, 1723c7198b9cSTheodore Ts'o {Opt_removed, "check=none"}, /* mount option from ext2/3 */ 1724c7198b9cSTheodore Ts'o {Opt_removed, "nocheck"}, /* mount option from ext2/3 */ 1725c7198b9cSTheodore Ts'o {Opt_removed, "reservation"}, /* mount option from ext2/3 */ 1726c7198b9cSTheodore Ts'o {Opt_removed, "noreservation"}, /* mount option from ext2/3 */ 1727c7198b9cSTheodore Ts'o {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */ 1728f3f12faaSJosef Bacik {Opt_err, NULL}, 1729ac27a0ecSDave Kleikamp }; 1730ac27a0ecSDave Kleikamp 1731617ba13bSMingming Cao static ext4_fsblk_t get_sb_block(void **data) 1732ac27a0ecSDave Kleikamp { 1733617ba13bSMingming Cao ext4_fsblk_t sb_block; 1734ac27a0ecSDave Kleikamp char *options = (char *) *data; 1735ac27a0ecSDave Kleikamp 1736ac27a0ecSDave Kleikamp if (!options || strncmp(options, "sb=", 3) != 0) 1737ac27a0ecSDave Kleikamp return 1; /* Default location */ 17380b8e58a1SAndreas Dilger 1739ac27a0ecSDave Kleikamp options += 3; 17400b8e58a1SAndreas Dilger /* TODO: use simple_strtoll with >32bit ext4 */ 1741ac27a0ecSDave Kleikamp sb_block = simple_strtoul(options, &options, 0); 1742ac27a0ecSDave Kleikamp if (*options && *options != ',') { 17434776004fSTheodore Ts'o printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n", 1744ac27a0ecSDave Kleikamp (char *) *data); 1745ac27a0ecSDave Kleikamp return 1; 1746ac27a0ecSDave Kleikamp } 1747ac27a0ecSDave Kleikamp if (*options == ',') 1748ac27a0ecSDave Kleikamp options++; 1749ac27a0ecSDave Kleikamp *data = (void *) options; 17500b8e58a1SAndreas Dilger 1751ac27a0ecSDave Kleikamp return sb_block; 1752ac27a0ecSDave Kleikamp } 1753ac27a0ecSDave Kleikamp 1754b3881f74STheodore Ts'o #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3)) 1755d6006186SEric Biggers static const char deprecated_msg[] = 1756d6006186SEric Biggers "Mount option \"%s\" will be removed by %s\n" 1757437ca0fdSDmitry Monakhov "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n"; 1758b3881f74STheodore Ts'o 175956c50f11SDmitry Monakhov #ifdef CONFIG_QUOTA 176056c50f11SDmitry Monakhov static int set_qf_name(struct super_block *sb, int qtype, substring_t *args) 176156c50f11SDmitry Monakhov { 176256c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 176333458eabSTheodore Ts'o char *qname, *old_qname = get_qf_name(sb, sbi, qtype); 176403dafb5fSChen Gang int ret = -1; 176556c50f11SDmitry Monakhov 176633458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && !old_qname) { 176756c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 176856c50f11SDmitry Monakhov "Cannot change journaled " 176956c50f11SDmitry Monakhov "quota options when quota turned on"); 177057f73c2cSTheodore Ts'o return -1; 177156c50f11SDmitry Monakhov } 1772e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 1773c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, "Journaled quota options " 1774c325a67cSTheodore Ts'o "ignored when QUOTA feature is enabled"); 1775c325a67cSTheodore Ts'o return 1; 1776262b4662SJan Kara } 177756c50f11SDmitry Monakhov qname = match_strdup(args); 177856c50f11SDmitry Monakhov if (!qname) { 177956c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 178056c50f11SDmitry Monakhov "Not enough memory for storing quotafile name"); 178157f73c2cSTheodore Ts'o return -1; 178256c50f11SDmitry Monakhov } 178333458eabSTheodore Ts'o if (old_qname) { 178433458eabSTheodore Ts'o if (strcmp(old_qname, qname) == 0) 178503dafb5fSChen Gang ret = 1; 178603dafb5fSChen Gang else 178756c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 178803dafb5fSChen Gang "%s quota file already specified", 178903dafb5fSChen Gang QTYPE2NAME(qtype)); 179003dafb5fSChen Gang goto errout; 179156c50f11SDmitry Monakhov } 179203dafb5fSChen Gang if (strchr(qname, '/')) { 179356c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, 179456c50f11SDmitry Monakhov "quotafile must be on filesystem root"); 179503dafb5fSChen Gang goto errout; 179656c50f11SDmitry Monakhov } 179733458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], qname); 1798fd8c37ecSTheodore Ts'o set_opt(sb, QUOTA); 179956c50f11SDmitry Monakhov return 1; 180003dafb5fSChen Gang errout: 180103dafb5fSChen Gang kfree(qname); 180203dafb5fSChen Gang return ret; 180356c50f11SDmitry Monakhov } 180456c50f11SDmitry Monakhov 180556c50f11SDmitry Monakhov static int clear_qf_name(struct super_block *sb, int qtype) 180656c50f11SDmitry Monakhov { 180756c50f11SDmitry Monakhov 180856c50f11SDmitry Monakhov struct ext4_sb_info *sbi = EXT4_SB(sb); 180933458eabSTheodore Ts'o char *old_qname = get_qf_name(sb, sbi, qtype); 181056c50f11SDmitry Monakhov 181133458eabSTheodore Ts'o if (sb_any_quota_loaded(sb) && old_qname) { 181256c50f11SDmitry Monakhov ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options" 181356c50f11SDmitry Monakhov " when quota turned on"); 181457f73c2cSTheodore Ts'o return -1; 181556c50f11SDmitry Monakhov } 181633458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[qtype], NULL); 181733458eabSTheodore Ts'o synchronize_rcu(); 181833458eabSTheodore Ts'o kfree(old_qname); 181956c50f11SDmitry Monakhov return 1; 182056c50f11SDmitry Monakhov } 182156c50f11SDmitry Monakhov #endif 182256c50f11SDmitry Monakhov 182326092bf5STheodore Ts'o #define MOPT_SET 0x0001 182426092bf5STheodore Ts'o #define MOPT_CLEAR 0x0002 182526092bf5STheodore Ts'o #define MOPT_NOSUPPORT 0x0004 182626092bf5STheodore Ts'o #define MOPT_EXPLICIT 0x0008 182726092bf5STheodore Ts'o #define MOPT_CLEAR_ERR 0x0010 182826092bf5STheodore Ts'o #define MOPT_GTE0 0x0020 182926092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 183026092bf5STheodore Ts'o #define MOPT_Q 0 183126092bf5STheodore Ts'o #define MOPT_QFMT 0x0040 183226092bf5STheodore Ts'o #else 183326092bf5STheodore Ts'o #define MOPT_Q MOPT_NOSUPPORT 183426092bf5STheodore Ts'o #define MOPT_QFMT MOPT_NOSUPPORT 183526092bf5STheodore Ts'o #endif 183626092bf5STheodore Ts'o #define MOPT_DATAJ 0x0080 18378dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT2 0x0100 18388dc0aa8cSTheodore Ts'o #define MOPT_NO_EXT3 0x0200 18398dc0aa8cSTheodore Ts'o #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3) 1840ad4eec61SEric Sandeen #define MOPT_STRING 0x0400 18419cb20f94SIra Weiny #define MOPT_SKIP 0x0800 184226092bf5STheodore Ts'o 184326092bf5STheodore Ts'o static const struct mount_opts { 184426092bf5STheodore Ts'o int token; 184526092bf5STheodore Ts'o int mount_opt; 184626092bf5STheodore Ts'o int flags; 184726092bf5STheodore Ts'o } ext4_mount_opts[] = { 184826092bf5STheodore Ts'o {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET}, 184926092bf5STheodore Ts'o {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR}, 185026092bf5STheodore Ts'o {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET}, 185126092bf5STheodore Ts'o {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR}, 185226092bf5STheodore Ts'o {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET}, 185326092bf5STheodore Ts'o {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR}, 18548dc0aa8cSTheodore Ts'o {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, 18558dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET}, 18568dc0aa8cSTheodore Ts'o {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, 18578dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 185826092bf5STheodore Ts'o {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET}, 185926092bf5STheodore Ts'o {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR}, 18608dc0aa8cSTheodore Ts'o {Opt_delalloc, EXT4_MOUNT_DELALLOC, 18618dc0aa8cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 18628dc0aa8cSTheodore Ts'o {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, 186359d9fa5cSTheodore Ts'o MOPT_EXT4_ONLY | MOPT_CLEAR}, 1864327eaf73STheodore Ts'o {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET}, 1865327eaf73STheodore Ts'o {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR}, 1866c6d3d56dSDarrick J. Wong {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 1867c6d3d56dSDarrick J. Wong MOPT_EXT4_ONLY | MOPT_CLEAR}, 18688dc0aa8cSTheodore Ts'o {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, 18691e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 187026092bf5STheodore Ts'o {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT | 18718dc0aa8cSTheodore Ts'o EXT4_MOUNT_JOURNAL_CHECKSUM), 18721e381f60SDmitry Monakhov MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT}, 18738dc0aa8cSTheodore Ts'o {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET}, 187426092bf5STheodore Ts'o {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR}, 187526092bf5STheodore Ts'o {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR}, 187626092bf5STheodore Ts'o {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR}, 18778dc0aa8cSTheodore Ts'o {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, 18787915a861SAles Novak MOPT_NO_EXT2}, 18798dc0aa8cSTheodore Ts'o {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, 18807915a861SAles Novak MOPT_NO_EXT2}, 188126092bf5STheodore Ts'o {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET}, 188226092bf5STheodore Ts'o {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR}, 188326092bf5STheodore Ts'o {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET}, 188426092bf5STheodore Ts'o {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR}, 188526092bf5STheodore Ts'o {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR}, 188626092bf5STheodore Ts'o {Opt_commit, 0, MOPT_GTE0}, 188726092bf5STheodore Ts'o {Opt_max_batch_time, 0, MOPT_GTE0}, 188826092bf5STheodore Ts'o {Opt_min_batch_time, 0, MOPT_GTE0}, 188926092bf5STheodore Ts'o {Opt_inode_readahead_blks, 0, MOPT_GTE0}, 189026092bf5STheodore Ts'o {Opt_init_itable, 0, MOPT_GTE0}, 18919cb20f94SIra Weiny {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP}, 18929cb20f94SIra Weiny {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS, 18939cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 18949cb20f94SIra Weiny {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE, 18959cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 18969cb20f94SIra Weiny {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER, 18979cb20f94SIra Weiny MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP}, 189826092bf5STheodore Ts'o {Opt_stripe, 0, MOPT_GTE0}, 18990efb3b23SJan Kara {Opt_resuid, 0, MOPT_GTE0}, 19000efb3b23SJan Kara {Opt_resgid, 0, MOPT_GTE0}, 19015ba92bcfSCarlos Maiolino {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 19025ba92bcfSCarlos Maiolino {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING}, 19035ba92bcfSCarlos Maiolino {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0}, 19048dc0aa8cSTheodore Ts'o {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 19058dc0aa8cSTheodore Ts'o {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ}, 19068dc0aa8cSTheodore Ts'o {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, 19078dc0aa8cSTheodore Ts'o MOPT_NO_EXT2 | MOPT_DATAJ}, 190826092bf5STheodore Ts'o {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET}, 190926092bf5STheodore Ts'o {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR}, 191026092bf5STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 191126092bf5STheodore Ts'o {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET}, 191226092bf5STheodore Ts'o {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR}, 191326092bf5STheodore Ts'o #else 191426092bf5STheodore Ts'o {Opt_acl, 0, MOPT_NOSUPPORT}, 191526092bf5STheodore Ts'o {Opt_noacl, 0, MOPT_NOSUPPORT}, 191626092bf5STheodore Ts'o #endif 191726092bf5STheodore Ts'o {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET}, 191826092bf5STheodore Ts'o {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET}, 1919670e9875STheodore Ts'o {Opt_debug_want_extra_isize, 0, MOPT_GTE0}, 192026092bf5STheodore Ts'o {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q}, 192126092bf5STheodore Ts'o {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, 192226092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 192326092bf5STheodore Ts'o {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA, 192426092bf5STheodore Ts'o MOPT_SET | MOPT_Q}, 192549da9392SJan Kara {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA, 192649da9392SJan Kara MOPT_SET | MOPT_Q}, 192726092bf5STheodore Ts'o {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA | 192849da9392SJan Kara EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA), 192949da9392SJan Kara MOPT_CLEAR | MOPT_Q}, 193026092bf5STheodore Ts'o {Opt_usrjquota, 0, MOPT_Q}, 193126092bf5STheodore Ts'o {Opt_grpjquota, 0, MOPT_Q}, 193226092bf5STheodore Ts'o {Opt_offusrjquota, 0, MOPT_Q}, 193326092bf5STheodore Ts'o {Opt_offgrpjquota, 0, MOPT_Q}, 193426092bf5STheodore Ts'o {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT}, 193526092bf5STheodore Ts'o {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT}, 193626092bf5STheodore Ts'o {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT}, 1937df981d03STheodore Ts'o {Opt_max_dir_size_kb, 0, MOPT_GTE0}, 1938ed318a6cSEric Biggers {Opt_test_dummy_encryption, 0, MOPT_STRING}, 1939cdb7ee4cSTahsin Erdogan {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET}, 19403d392b26STheodore Ts'o {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS, 19413d392b26STheodore Ts'o MOPT_SET}, 194226092bf5STheodore Ts'o {Opt_err, 0, 0} 194326092bf5STheodore Ts'o }; 194426092bf5STheodore Ts'o 1945c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 1946c83ad55eSGabriel Krisman Bertazi static const struct ext4_sb_encodings { 1947c83ad55eSGabriel Krisman Bertazi __u16 magic; 1948c83ad55eSGabriel Krisman Bertazi char *name; 1949c83ad55eSGabriel Krisman Bertazi char *version; 1950c83ad55eSGabriel Krisman Bertazi } ext4_sb_encoding_map[] = { 1951c83ad55eSGabriel Krisman Bertazi {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"}, 1952c83ad55eSGabriel Krisman Bertazi }; 1953c83ad55eSGabriel Krisman Bertazi 1954c83ad55eSGabriel Krisman Bertazi static int ext4_sb_read_encoding(const struct ext4_super_block *es, 1955c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings **encoding, 1956c83ad55eSGabriel Krisman Bertazi __u16 *flags) 1957c83ad55eSGabriel Krisman Bertazi { 1958c83ad55eSGabriel Krisman Bertazi __u16 magic = le16_to_cpu(es->s_encoding); 1959c83ad55eSGabriel Krisman Bertazi int i; 1960c83ad55eSGabriel Krisman Bertazi 1961c83ad55eSGabriel Krisman Bertazi for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++) 1962c83ad55eSGabriel Krisman Bertazi if (magic == ext4_sb_encoding_map[i].magic) 1963c83ad55eSGabriel Krisman Bertazi break; 1964c83ad55eSGabriel Krisman Bertazi 1965c83ad55eSGabriel Krisman Bertazi if (i >= ARRAY_SIZE(ext4_sb_encoding_map)) 1966c83ad55eSGabriel Krisman Bertazi return -EINVAL; 1967c83ad55eSGabriel Krisman Bertazi 1968c83ad55eSGabriel Krisman Bertazi *encoding = &ext4_sb_encoding_map[i]; 1969c83ad55eSGabriel Krisman Bertazi *flags = le16_to_cpu(es->s_encoding_flags); 1970c83ad55eSGabriel Krisman Bertazi 1971c83ad55eSGabriel Krisman Bertazi return 0; 1972c83ad55eSGabriel Krisman Bertazi } 1973c83ad55eSGabriel Krisman Bertazi #endif 1974c83ad55eSGabriel Krisman Bertazi 1975ed318a6cSEric Biggers static int ext4_set_test_dummy_encryption(struct super_block *sb, 1976ed318a6cSEric Biggers const char *opt, 1977ed318a6cSEric Biggers const substring_t *arg, 1978ed318a6cSEric Biggers bool is_remount) 1979ed318a6cSEric Biggers { 1980ed318a6cSEric Biggers #ifdef CONFIG_FS_ENCRYPTION 1981ed318a6cSEric Biggers struct ext4_sb_info *sbi = EXT4_SB(sb); 1982ed318a6cSEric Biggers int err; 1983ed318a6cSEric Biggers 1984ed318a6cSEric Biggers /* 1985ed318a6cSEric Biggers * This mount option is just for testing, and it's not worthwhile to 1986ed318a6cSEric Biggers * implement the extra complexity (e.g. RCU protection) that would be 1987ed318a6cSEric Biggers * needed to allow it to be set or changed during remount. We do allow 1988ed318a6cSEric Biggers * it to be specified during remount, but only if there is no change. 1989ed318a6cSEric Biggers */ 1990ed318a6cSEric Biggers if (is_remount && !sbi->s_dummy_enc_ctx.ctx) { 1991ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1992ed318a6cSEric Biggers "Can't set test_dummy_encryption on remount"); 1993ed318a6cSEric Biggers return -1; 1994ed318a6cSEric Biggers } 1995ed318a6cSEric Biggers err = fscrypt_set_test_dummy_encryption(sb, arg, &sbi->s_dummy_enc_ctx); 1996ed318a6cSEric Biggers if (err) { 1997ed318a6cSEric Biggers if (err == -EEXIST) 1998ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 1999ed318a6cSEric Biggers "Can't change test_dummy_encryption on remount"); 2000ed318a6cSEric Biggers else if (err == -EINVAL) 2001ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2002ed318a6cSEric Biggers "Value of option \"%s\" is unrecognized", opt); 2003ed318a6cSEric Biggers else 2004ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2005ed318a6cSEric Biggers "Error processing option \"%s\" [%d]", 2006ed318a6cSEric Biggers opt, err); 2007ed318a6cSEric Biggers return -1; 2008ed318a6cSEric Biggers } 2009ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled"); 2010ed318a6cSEric Biggers #else 2011ed318a6cSEric Biggers ext4_msg(sb, KERN_WARNING, 2012ed318a6cSEric Biggers "Test dummy encryption mount option ignored"); 2013ed318a6cSEric Biggers #endif 2014ed318a6cSEric Biggers return 1; 2015ed318a6cSEric Biggers } 2016ed318a6cSEric Biggers 201726092bf5STheodore Ts'o static int handle_mount_opt(struct super_block *sb, char *opt, int token, 201826092bf5STheodore Ts'o substring_t *args, unsigned long *journal_devnum, 201926092bf5STheodore Ts'o unsigned int *journal_ioprio, int is_remount) 202026092bf5STheodore Ts'o { 202126092bf5STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 202226092bf5STheodore Ts'o const struct mount_opts *m; 202308cefc7aSEric W. Biederman kuid_t uid; 202408cefc7aSEric W. Biederman kgid_t gid; 202526092bf5STheodore Ts'o int arg = 0; 202626092bf5STheodore Ts'o 202757f73c2cSTheodore Ts'o #ifdef CONFIG_QUOTA 202857f73c2cSTheodore Ts'o if (token == Opt_usrjquota) 202957f73c2cSTheodore Ts'o return set_qf_name(sb, USRQUOTA, &args[0]); 203057f73c2cSTheodore Ts'o else if (token == Opt_grpjquota) 203157f73c2cSTheodore Ts'o return set_qf_name(sb, GRPQUOTA, &args[0]); 203257f73c2cSTheodore Ts'o else if (token == Opt_offusrjquota) 203357f73c2cSTheodore Ts'o return clear_qf_name(sb, USRQUOTA); 203457f73c2cSTheodore Ts'o else if (token == Opt_offgrpjquota) 203557f73c2cSTheodore Ts'o return clear_qf_name(sb, GRPQUOTA); 203657f73c2cSTheodore Ts'o #endif 203726092bf5STheodore Ts'o switch (token) { 2038f7048605STheodore Ts'o case Opt_noacl: 2039f7048605STheodore Ts'o case Opt_nouser_xattr: 2040f7048605STheodore Ts'o ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5"); 2041f7048605STheodore Ts'o break; 204226092bf5STheodore Ts'o case Opt_sb: 204326092bf5STheodore Ts'o return 1; /* handled by get_sb_block() */ 204426092bf5STheodore Ts'o case Opt_removed: 20455f3633e3SJan Kara ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt); 204626092bf5STheodore Ts'o return 1; 204726092bf5STheodore Ts'o case Opt_abort: 204826092bf5STheodore Ts'o sbi->s_mount_flags |= EXT4_MF_FS_ABORTED; 204926092bf5STheodore Ts'o return 1; 205026092bf5STheodore Ts'o case Opt_i_version: 2051357fdad0SMatthew Garrett sb->s_flags |= SB_I_VERSION; 205226092bf5STheodore Ts'o return 1; 2053a26f4992STheodore Ts'o case Opt_lazytime: 20541751e8a6SLinus Torvalds sb->s_flags |= SB_LAZYTIME; 2055a26f4992STheodore Ts'o return 1; 2056a26f4992STheodore Ts'o case Opt_nolazytime: 20571751e8a6SLinus Torvalds sb->s_flags &= ~SB_LAZYTIME; 2058a26f4992STheodore Ts'o return 1; 20594f74d15fSEric Biggers case Opt_inlinecrypt: 20604f74d15fSEric Biggers #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT 20614f74d15fSEric Biggers sb->s_flags |= SB_INLINECRYPT; 20624f74d15fSEric Biggers #else 20634f74d15fSEric Biggers ext4_msg(sb, KERN_ERR, "inline encryption not supported"); 20644f74d15fSEric Biggers #endif 20654f74d15fSEric Biggers return 1; 206626092bf5STheodore Ts'o } 206726092bf5STheodore Ts'o 20685f3633e3SJan Kara for (m = ext4_mount_opts; m->token != Opt_err; m++) 20695f3633e3SJan Kara if (token == m->token) 20705f3633e3SJan Kara break; 20715f3633e3SJan Kara 20725f3633e3SJan Kara if (m->token == Opt_err) { 20735f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" " 20745f3633e3SJan Kara "or missing value", opt); 20755f3633e3SJan Kara return -1; 20765f3633e3SJan Kara } 20775f3633e3SJan Kara 20788dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) { 20798dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 20808dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext2", opt); 20818dc0aa8cSTheodore Ts'o return -1; 20828dc0aa8cSTheodore Ts'o } 20838dc0aa8cSTheodore Ts'o if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) { 20848dc0aa8cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 20858dc0aa8cSTheodore Ts'o "Mount option \"%s\" incompatible with ext3", opt); 20868dc0aa8cSTheodore Ts'o return -1; 20878dc0aa8cSTheodore Ts'o } 20888dc0aa8cSTheodore Ts'o 2089ad4eec61SEric Sandeen if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg)) 20900efb3b23SJan Kara return -1; 209126092bf5STheodore Ts'o if (args->from && (m->flags & MOPT_GTE0) && (arg < 0)) 209226092bf5STheodore Ts'o return -1; 2093c93cf2d7SDmitry Monakhov if (m->flags & MOPT_EXPLICIT) { 2094c93cf2d7SDmitry Monakhov if (m->mount_opt & EXT4_MOUNT_DELALLOC) { 209526092bf5STheodore Ts'o set_opt2(sb, EXPLICIT_DELALLOC); 20961e381f60SDmitry Monakhov } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) { 20971e381f60SDmitry Monakhov set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM); 2098c93cf2d7SDmitry Monakhov } else 2099c93cf2d7SDmitry Monakhov return -1; 2100c93cf2d7SDmitry Monakhov } 210126092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR_ERR) 210226092bf5STheodore Ts'o clear_opt(sb, ERRORS_MASK); 210326092bf5STheodore Ts'o if (token == Opt_noquota && sb_any_quota_loaded(sb)) { 210426092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "Cannot change quota " 210526092bf5STheodore Ts'o "options when quota turned on"); 210626092bf5STheodore Ts'o return -1; 210726092bf5STheodore Ts'o } 210826092bf5STheodore Ts'o 210926092bf5STheodore Ts'o if (m->flags & MOPT_NOSUPPORT) { 211026092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "%s option not supported", opt); 211126092bf5STheodore Ts'o } else if (token == Opt_commit) { 211226092bf5STheodore Ts'o if (arg == 0) 211326092bf5STheodore Ts'o arg = JBD2_DEFAULT_MAX_COMMIT_AGE; 21149ba55543Szhangyi (F) else if (arg > INT_MAX / HZ) { 21159ba55543Szhangyi (F) ext4_msg(sb, KERN_ERR, 21169ba55543Szhangyi (F) "Invalid commit interval %d, " 21179ba55543Szhangyi (F) "must be smaller than %d", 21189ba55543Szhangyi (F) arg, INT_MAX / HZ); 21199ba55543Szhangyi (F) return -1; 21209ba55543Szhangyi (F) } 212126092bf5STheodore Ts'o sbi->s_commit_interval = HZ * arg; 2122670e9875STheodore Ts'o } else if (token == Opt_debug_want_extra_isize) { 21239803387cSTheodore Ts'o if ((arg & 1) || 21249803387cSTheodore Ts'o (arg < 4) || 21259803387cSTheodore Ts'o (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) { 21269803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 21279803387cSTheodore Ts'o "Invalid want_extra_isize %d", arg); 21289803387cSTheodore Ts'o return -1; 21299803387cSTheodore Ts'o } 2130670e9875STheodore Ts'o sbi->s_want_extra_isize = arg; 213126092bf5STheodore Ts'o } else if (token == Opt_max_batch_time) { 213226092bf5STheodore Ts'o sbi->s_max_batch_time = arg; 213326092bf5STheodore Ts'o } else if (token == Opt_min_batch_time) { 213426092bf5STheodore Ts'o sbi->s_min_batch_time = arg; 213526092bf5STheodore Ts'o } else if (token == Opt_inode_readahead_blks) { 2136e33e60eaSJan Kara if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) { 2137e33e60eaSJan Kara ext4_msg(sb, KERN_ERR, 2138e33e60eaSJan Kara "EXT4-fs: inode_readahead_blks must be " 2139e33e60eaSJan Kara "0 or a power of 2 smaller than 2^31"); 214026092bf5STheodore Ts'o return -1; 214126092bf5STheodore Ts'o } 214226092bf5STheodore Ts'o sbi->s_inode_readahead_blks = arg; 214326092bf5STheodore Ts'o } else if (token == Opt_init_itable) { 214426092bf5STheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 214526092bf5STheodore Ts'o if (!args->from) 214626092bf5STheodore Ts'o arg = EXT4_DEF_LI_WAIT_MULT; 214726092bf5STheodore Ts'o sbi->s_li_wait_mult = arg; 2148df981d03STheodore Ts'o } else if (token == Opt_max_dir_size_kb) { 2149df981d03STheodore Ts'o sbi->s_max_dir_size_kb = arg; 215026092bf5STheodore Ts'o } else if (token == Opt_stripe) { 215126092bf5STheodore Ts'o sbi->s_stripe = arg; 21520efb3b23SJan Kara } else if (token == Opt_resuid) { 21530efb3b23SJan Kara uid = make_kuid(current_user_ns(), arg); 21540efb3b23SJan Kara if (!uid_valid(uid)) { 21555f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg); 21560efb3b23SJan Kara return -1; 21570efb3b23SJan Kara } 21580efb3b23SJan Kara sbi->s_resuid = uid; 21590efb3b23SJan Kara } else if (token == Opt_resgid) { 21600efb3b23SJan Kara gid = make_kgid(current_user_ns(), arg); 21610efb3b23SJan Kara if (!gid_valid(gid)) { 21625f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg); 21630efb3b23SJan Kara return -1; 21640efb3b23SJan Kara } 21650efb3b23SJan Kara sbi->s_resgid = gid; 21660efb3b23SJan Kara } else if (token == Opt_journal_dev) { 21670efb3b23SJan Kara if (is_remount) { 21680efb3b23SJan Kara ext4_msg(sb, KERN_ERR, 21690efb3b23SJan Kara "Cannot specify journal on remount"); 21700efb3b23SJan Kara return -1; 21710efb3b23SJan Kara } 21720efb3b23SJan Kara *journal_devnum = arg; 2173ad4eec61SEric Sandeen } else if (token == Opt_journal_path) { 2174ad4eec61SEric Sandeen char *journal_path; 2175ad4eec61SEric Sandeen struct inode *journal_inode; 2176ad4eec61SEric Sandeen struct path path; 2177ad4eec61SEric Sandeen int error; 2178ad4eec61SEric Sandeen 2179ad4eec61SEric Sandeen if (is_remount) { 2180ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, 2181ad4eec61SEric Sandeen "Cannot specify journal on remount"); 2182ad4eec61SEric Sandeen return -1; 2183ad4eec61SEric Sandeen } 2184ad4eec61SEric Sandeen journal_path = match_strdup(&args[0]); 2185ad4eec61SEric Sandeen if (!journal_path) { 2186ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not dup " 2187ad4eec61SEric Sandeen "journal device string"); 2188ad4eec61SEric Sandeen return -1; 2189ad4eec61SEric Sandeen } 2190ad4eec61SEric Sandeen 2191ad4eec61SEric Sandeen error = kern_path(journal_path, LOOKUP_FOLLOW, &path); 2192ad4eec61SEric Sandeen if (error) { 2193ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: could not find " 2194ad4eec61SEric Sandeen "journal device path: error %d", error); 2195ad4eec61SEric Sandeen kfree(journal_path); 2196ad4eec61SEric Sandeen return -1; 2197ad4eec61SEric Sandeen } 2198ad4eec61SEric Sandeen 21992b0143b5SDavid Howells journal_inode = d_inode(path.dentry); 2200ad4eec61SEric Sandeen if (!S_ISBLK(journal_inode->i_mode)) { 2201ad4eec61SEric Sandeen ext4_msg(sb, KERN_ERR, "error: journal path %s " 2202ad4eec61SEric Sandeen "is not a block device", journal_path); 2203ad4eec61SEric Sandeen path_put(&path); 2204ad4eec61SEric Sandeen kfree(journal_path); 2205ad4eec61SEric Sandeen return -1; 2206ad4eec61SEric Sandeen } 2207ad4eec61SEric Sandeen 2208ad4eec61SEric Sandeen *journal_devnum = new_encode_dev(journal_inode->i_rdev); 2209ad4eec61SEric Sandeen path_put(&path); 2210ad4eec61SEric Sandeen kfree(journal_path); 22110efb3b23SJan Kara } else if (token == Opt_journal_ioprio) { 22120efb3b23SJan Kara if (arg > 7) { 22135f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Invalid journal IO priority" 22140efb3b23SJan Kara " (must be 0-7)"); 22150efb3b23SJan Kara return -1; 22160efb3b23SJan Kara } 22170efb3b23SJan Kara *journal_ioprio = 22180efb3b23SJan Kara IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg); 22196ddb2447STheodore Ts'o } else if (token == Opt_test_dummy_encryption) { 2220ed318a6cSEric Biggers return ext4_set_test_dummy_encryption(sb, opt, &args[0], 2221ed318a6cSEric Biggers is_remount); 222226092bf5STheodore Ts'o } else if (m->flags & MOPT_DATAJ) { 222326092bf5STheodore Ts'o if (is_remount) { 222426092bf5STheodore Ts'o if (!sbi->s_journal) 222526092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option"); 22265f3633e3SJan Kara else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) { 222726092bf5STheodore Ts'o ext4_msg(sb, KERN_ERR, 222826092bf5STheodore Ts'o "Cannot change data mode on remount"); 222926092bf5STheodore Ts'o return -1; 223026092bf5STheodore Ts'o } 223126092bf5STheodore Ts'o } else { 223226092bf5STheodore Ts'o clear_opt(sb, DATA_FLAGS); 223326092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 223426092bf5STheodore Ts'o } 223526092bf5STheodore Ts'o #ifdef CONFIG_QUOTA 223626092bf5STheodore Ts'o } else if (m->flags & MOPT_QFMT) { 223726092bf5STheodore Ts'o if (sb_any_quota_loaded(sb) && 223826092bf5STheodore Ts'o sbi->s_jquota_fmt != m->mount_opt) { 22395f3633e3SJan Kara ext4_msg(sb, KERN_ERR, "Cannot change journaled " 22405f3633e3SJan Kara "quota options when quota turned on"); 224126092bf5STheodore Ts'o return -1; 224226092bf5STheodore Ts'o } 2243e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) { 2244c325a67cSTheodore Ts'o ext4_msg(sb, KERN_INFO, 2245c325a67cSTheodore Ts'o "Quota format mount options ignored " 2246262b4662SJan Kara "when QUOTA feature is enabled"); 2247c325a67cSTheodore Ts'o return 1; 2248262b4662SJan Kara } 224926092bf5STheodore Ts'o sbi->s_jquota_fmt = m->mount_opt; 225026092bf5STheodore Ts'o #endif 22519cb20f94SIra Weiny } else if (token == Opt_dax || token == Opt_dax_always || 22529cb20f94SIra Weiny token == Opt_dax_inode || token == Opt_dax_never) { 2253ef83b6e8SDan Williams #ifdef CONFIG_FS_DAX 22549cb20f94SIra Weiny switch (token) { 22559cb20f94SIra Weiny case Opt_dax: 22569cb20f94SIra Weiny case Opt_dax_always: 2257829b37b8STheodore Ts'o if (is_remount && 2258829b37b8STheodore Ts'o (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2259829b37b8STheodore Ts'o (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) { 2260829b37b8STheodore Ts'o fail_dax_change_remount: 2261829b37b8STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't change " 2262829b37b8STheodore Ts'o "dax mount option while remounting"); 2263829b37b8STheodore Ts'o return -1; 2264829b37b8STheodore Ts'o } 2265829b37b8STheodore Ts'o if (is_remount && 2266829b37b8STheodore Ts'o (test_opt(sb, DATA_FLAGS) == 2267829b37b8STheodore Ts'o EXT4_MOUNT_JOURNAL_DATA)) { 2268829b37b8STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 2269829b37b8STheodore Ts'o "both data=journal and dax"); 2270829b37b8STheodore Ts'o return -1; 2271829b37b8STheodore Ts'o } 2272ef83b6e8SDan Williams ext4_msg(sb, KERN_WARNING, 2273ef83b6e8SDan Williams "DAX enabled. Warning: EXPERIMENTAL, use at your own risk"); 22749cb20f94SIra Weiny sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS; 22759cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 22769cb20f94SIra Weiny break; 22779cb20f94SIra Weiny case Opt_dax_never: 2278829b37b8STheodore Ts'o if (is_remount && 2279829b37b8STheodore Ts'o (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2280829b37b8STheodore Ts'o (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) 2281829b37b8STheodore Ts'o goto fail_dax_change_remount; 22829cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 22839cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 22849cb20f94SIra Weiny break; 22859cb20f94SIra Weiny case Opt_dax_inode: 2286829b37b8STheodore Ts'o if (is_remount && 2287829b37b8STheodore Ts'o ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) || 2288829b37b8STheodore Ts'o (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) || 2289829b37b8STheodore Ts'o !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) 2290829b37b8STheodore Ts'o goto fail_dax_change_remount; 22919cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 22929cb20f94SIra Weiny sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER; 22939cb20f94SIra Weiny /* Strictly for printing options */ 22949cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE; 22959cb20f94SIra Weiny break; 22969cb20f94SIra Weiny } 2297ef83b6e8SDan Williams #else 2298923ae0ffSRoss Zwisler ext4_msg(sb, KERN_INFO, "dax option not supported"); 22999cb20f94SIra Weiny sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER; 23009cb20f94SIra Weiny sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS; 2301923ae0ffSRoss Zwisler return -1; 2302923ae0ffSRoss Zwisler #endif 23037915a861SAles Novak } else if (token == Opt_data_err_abort) { 23047915a861SAles Novak sbi->s_mount_opt |= m->mount_opt; 23057915a861SAles Novak } else if (token == Opt_data_err_ignore) { 23067915a861SAles Novak sbi->s_mount_opt &= ~m->mount_opt; 230726092bf5STheodore Ts'o } else { 230826092bf5STheodore Ts'o if (!args->from) 230926092bf5STheodore Ts'o arg = 1; 231026092bf5STheodore Ts'o if (m->flags & MOPT_CLEAR) 231126092bf5STheodore Ts'o arg = !arg; 231226092bf5STheodore Ts'o else if (unlikely(!(m->flags & MOPT_SET))) { 231326092bf5STheodore Ts'o ext4_msg(sb, KERN_WARNING, 231426092bf5STheodore Ts'o "buggy handling of option %s", opt); 231526092bf5STheodore Ts'o WARN_ON(1); 231626092bf5STheodore Ts'o return -1; 231726092bf5STheodore Ts'o } 231826092bf5STheodore Ts'o if (arg != 0) 231926092bf5STheodore Ts'o sbi->s_mount_opt |= m->mount_opt; 232026092bf5STheodore Ts'o else 232126092bf5STheodore Ts'o sbi->s_mount_opt &= ~m->mount_opt; 232226092bf5STheodore Ts'o } 232326092bf5STheodore Ts'o return 1; 232426092bf5STheodore Ts'o } 232526092bf5STheodore Ts'o 2326ac27a0ecSDave Kleikamp static int parse_options(char *options, struct super_block *sb, 2327c3191067STheodore Ts'o unsigned long *journal_devnum, 2328b3881f74STheodore Ts'o unsigned int *journal_ioprio, 2329661aa520SEric Sandeen int is_remount) 2330ac27a0ecSDave Kleikamp { 23311e1a76edSOlof Johansson struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb); 233233458eabSTheodore Ts'o char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name; 2333ac27a0ecSDave Kleikamp substring_t args[MAX_OPT_ARGS]; 233426092bf5STheodore Ts'o int token; 2335ac27a0ecSDave Kleikamp 2336ac27a0ecSDave Kleikamp if (!options) 2337ac27a0ecSDave Kleikamp return 1; 2338ac27a0ecSDave Kleikamp 2339ac27a0ecSDave Kleikamp while ((p = strsep(&options, ",")) != NULL) { 2340ac27a0ecSDave Kleikamp if (!*p) 2341ac27a0ecSDave Kleikamp continue; 234215121c18SEric Sandeen /* 234315121c18SEric Sandeen * Initialize args struct so we know whether arg was 234415121c18SEric Sandeen * found; some options take optional arguments. 234515121c18SEric Sandeen */ 2346caecd0afSSachin Kamat args[0].to = args[0].from = NULL; 2347ac27a0ecSDave Kleikamp token = match_token(p, tokens, args); 234826092bf5STheodore Ts'o if (handle_mount_opt(sb, p, token, args, journal_devnum, 234926092bf5STheodore Ts'o journal_ioprio, is_remount) < 0) 2350ac27a0ecSDave Kleikamp return 0; 2351ac27a0ecSDave Kleikamp } 2352ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 235349da9392SJan Kara /* 235449da9392SJan Kara * We do the test below only for project quotas. 'usrquota' and 235549da9392SJan Kara * 'grpquota' mount options are allowed even without quota feature 235649da9392SJan Kara * to support legacy quotas in quota files. 235749da9392SJan Kara */ 235849da9392SJan Kara if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) { 235949da9392SJan Kara ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. " 236049da9392SJan Kara "Cannot enable project quota enforcement."); 236149da9392SJan Kara return 0; 236249da9392SJan Kara } 236333458eabSTheodore Ts'o usr_qf_name = get_qf_name(sb, sbi, USRQUOTA); 236433458eabSTheodore Ts'o grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA); 236533458eabSTheodore Ts'o if (usr_qf_name || grp_qf_name) { 236633458eabSTheodore Ts'o if (test_opt(sb, USRQUOTA) && usr_qf_name) 2367fd8c37ecSTheodore Ts'o clear_opt(sb, USRQUOTA); 2368ac27a0ecSDave Kleikamp 236933458eabSTheodore Ts'o if (test_opt(sb, GRPQUOTA) && grp_qf_name) 2370fd8c37ecSTheodore Ts'o clear_opt(sb, GRPQUOTA); 2371ac27a0ecSDave Kleikamp 237256c50f11SDmitry Monakhov if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) { 2373b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "old and new quota " 2374b31e1552SEric Sandeen "format mixing"); 2375ac27a0ecSDave Kleikamp return 0; 2376ac27a0ecSDave Kleikamp } 2377ac27a0ecSDave Kleikamp 2378ac27a0ecSDave Kleikamp if (!sbi->s_jquota_fmt) { 2379b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journaled quota format " 2380b31e1552SEric Sandeen "not specified"); 2381ac27a0ecSDave Kleikamp return 0; 2382ac27a0ecSDave Kleikamp } 2383ac27a0ecSDave Kleikamp } 2384ac27a0ecSDave Kleikamp #endif 2385626b035bSRitesh Harjani if (test_opt(sb, DIOREAD_NOLOCK)) { 2386626b035bSRitesh Harjani int blocksize = 2387626b035bSRitesh Harjani BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size); 2388626b035bSRitesh Harjani if (blocksize < PAGE_SIZE) 2389626b035bSRitesh Harjani ext4_msg(sb, KERN_WARNING, "Warning: mounting with an " 2390626b035bSRitesh Harjani "experimental mount option 'dioread_nolock' " 2391626b035bSRitesh Harjani "for blocksize < PAGE_SIZE"); 2392626b035bSRitesh Harjani } 2393ac27a0ecSDave Kleikamp return 1; 2394ac27a0ecSDave Kleikamp } 2395ac27a0ecSDave Kleikamp 23962adf6da8STheodore Ts'o static inline void ext4_show_quota_options(struct seq_file *seq, 23972adf6da8STheodore Ts'o struct super_block *sb) 23982adf6da8STheodore Ts'o { 23992adf6da8STheodore Ts'o #if defined(CONFIG_QUOTA) 24002adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 240133458eabSTheodore Ts'o char *usr_qf_name, *grp_qf_name; 24022adf6da8STheodore Ts'o 24032adf6da8STheodore Ts'o if (sbi->s_jquota_fmt) { 24042adf6da8STheodore Ts'o char *fmtname = ""; 24052adf6da8STheodore Ts'o 24062adf6da8STheodore Ts'o switch (sbi->s_jquota_fmt) { 24072adf6da8STheodore Ts'o case QFMT_VFS_OLD: 24082adf6da8STheodore Ts'o fmtname = "vfsold"; 24092adf6da8STheodore Ts'o break; 24102adf6da8STheodore Ts'o case QFMT_VFS_V0: 24112adf6da8STheodore Ts'o fmtname = "vfsv0"; 24122adf6da8STheodore Ts'o break; 24132adf6da8STheodore Ts'o case QFMT_VFS_V1: 24142adf6da8STheodore Ts'o fmtname = "vfsv1"; 24152adf6da8STheodore Ts'o break; 24162adf6da8STheodore Ts'o } 24172adf6da8STheodore Ts'o seq_printf(seq, ",jqfmt=%s", fmtname); 24182adf6da8STheodore Ts'o } 24192adf6da8STheodore Ts'o 242033458eabSTheodore Ts'o rcu_read_lock(); 242133458eabSTheodore Ts'o usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]); 242233458eabSTheodore Ts'o grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]); 242333458eabSTheodore Ts'o if (usr_qf_name) 242433458eabSTheodore Ts'o seq_show_option(seq, "usrjquota", usr_qf_name); 242533458eabSTheodore Ts'o if (grp_qf_name) 242633458eabSTheodore Ts'o seq_show_option(seq, "grpjquota", grp_qf_name); 242733458eabSTheodore Ts'o rcu_read_unlock(); 24282adf6da8STheodore Ts'o #endif 24292adf6da8STheodore Ts'o } 24302adf6da8STheodore Ts'o 24315a916be1STheodore Ts'o static const char *token2str(int token) 24325a916be1STheodore Ts'o { 243350df9fd5SHerton Ronaldo Krzesinski const struct match_token *t; 24345a916be1STheodore Ts'o 24355a916be1STheodore Ts'o for (t = tokens; t->token != Opt_err; t++) 24365a916be1STheodore Ts'o if (t->token == token && !strchr(t->pattern, '=')) 24375a916be1STheodore Ts'o break; 24385a916be1STheodore Ts'o return t->pattern; 24395a916be1STheodore Ts'o } 24405a916be1STheodore Ts'o 24412adf6da8STheodore Ts'o /* 24422adf6da8STheodore Ts'o * Show an option if 24432adf6da8STheodore Ts'o * - it's set to a non-default value OR 24442adf6da8STheodore Ts'o * - if the per-sb default is different from the global default 24452adf6da8STheodore Ts'o */ 244666acdcf4STheodore Ts'o static int _ext4_show_options(struct seq_file *seq, struct super_block *sb, 244766acdcf4STheodore Ts'o int nodefs) 24482adf6da8STheodore Ts'o { 24492adf6da8STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 24502adf6da8STheodore Ts'o struct ext4_super_block *es = sbi->s_es; 245168afa7e0STyson Nottingham int def_errors, def_mount_opt = sbi->s_def_mount_opt; 24525a916be1STheodore Ts'o const struct mount_opts *m; 245366acdcf4STheodore Ts'o char sep = nodefs ? '\n' : ','; 24542adf6da8STheodore Ts'o 245566acdcf4STheodore Ts'o #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep) 245666acdcf4STheodore Ts'o #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg) 24572adf6da8STheodore Ts'o 24582adf6da8STheodore Ts'o if (sbi->s_sb_block != 1) 24595a916be1STheodore Ts'o SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block); 24605a916be1STheodore Ts'o 24615a916be1STheodore Ts'o for (m = ext4_mount_opts; m->token != Opt_err; m++) { 24625a916be1STheodore Ts'o int want_set = m->flags & MOPT_SET; 24635a916be1STheodore Ts'o if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) || 24649cb20f94SIra Weiny (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP) 24655a916be1STheodore Ts'o continue; 246668afa7e0STyson Nottingham if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt))) 24675a916be1STheodore Ts'o continue; /* skip if same as the default */ 24685a916be1STheodore Ts'o if ((want_set && 24695a916be1STheodore Ts'o (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) || 24705a916be1STheodore Ts'o (!want_set && (sbi->s_mount_opt & m->mount_opt))) 24715a916be1STheodore Ts'o continue; /* select Opt_noFoo vs Opt_Foo */ 24725a916be1STheodore Ts'o SEQ_OPTS_PRINT("%s", token2str(m->token)); 24735a916be1STheodore Ts'o } 24745a916be1STheodore Ts'o 247508cefc7aSEric W. Biederman if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) || 24765a916be1STheodore Ts'o le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) 247708cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resuid=%u", 247808cefc7aSEric W. Biederman from_kuid_munged(&init_user_ns, sbi->s_resuid)); 247908cefc7aSEric W. Biederman if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) || 24805a916be1STheodore Ts'o le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) 248108cefc7aSEric W. Biederman SEQ_OPTS_PRINT("resgid=%u", 248208cefc7aSEric W. Biederman from_kgid_munged(&init_user_ns, sbi->s_resgid)); 248366acdcf4STheodore Ts'o def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors); 24845a916be1STheodore Ts'o if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO) 24855a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=remount-ro"); 24862adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE) 24875a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=continue"); 24882adf6da8STheodore Ts'o if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC) 24895a916be1STheodore Ts'o SEQ_OPTS_PUTS("errors=panic"); 249066acdcf4STheodore Ts'o if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) 24915a916be1STheodore Ts'o SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ); 249266acdcf4STheodore Ts'o if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) 24935a916be1STheodore Ts'o SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time); 249466acdcf4STheodore Ts'o if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) 24955a916be1STheodore Ts'o SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time); 2496357fdad0SMatthew Garrett if (sb->s_flags & SB_I_VERSION) 24975a916be1STheodore Ts'o SEQ_OPTS_PUTS("i_version"); 249866acdcf4STheodore Ts'o if (nodefs || sbi->s_stripe) 24995a916be1STheodore Ts'o SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe); 250068afa7e0STyson Nottingham if (nodefs || EXT4_MOUNT_DATA_FLAGS & 250168afa7e0STyson Nottingham (sbi->s_mount_opt ^ def_mount_opt)) { 25022adf6da8STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 25035a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=journal"); 25042adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 25055a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=ordered"); 25062adf6da8STheodore Ts'o else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA) 25075a916be1STheodore Ts'o SEQ_OPTS_PUTS("data=writeback"); 25085a916be1STheodore Ts'o } 250966acdcf4STheodore Ts'o if (nodefs || 251066acdcf4STheodore Ts'o sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS) 25115a916be1STheodore Ts'o SEQ_OPTS_PRINT("inode_readahead_blks=%u", 25122adf6da8STheodore Ts'o sbi->s_inode_readahead_blks); 25132adf6da8STheodore Ts'o 2514ceec0376STyson Nottingham if (test_opt(sb, INIT_INODE_TABLE) && (nodefs || 251566acdcf4STheodore Ts'o (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT))) 25165a916be1STheodore Ts'o SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult); 2517df981d03STheodore Ts'o if (nodefs || sbi->s_max_dir_size_kb) 2518df981d03STheodore Ts'o SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb); 25197915a861SAles Novak if (test_opt(sb, DATA_ERR_ABORT)) 25207915a861SAles Novak SEQ_OPTS_PUTS("data_err=abort"); 2521ed318a6cSEric Biggers 2522ed318a6cSEric Biggers fscrypt_show_test_dummy_encryption(seq, sep, sb); 25232adf6da8STheodore Ts'o 25244f74d15fSEric Biggers if (sb->s_flags & SB_INLINECRYPT) 25254f74d15fSEric Biggers SEQ_OPTS_PUTS("inlinecrypt"); 25264f74d15fSEric Biggers 25279cb20f94SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 25289cb20f94SIra Weiny if (IS_EXT2_SB(sb)) 25299cb20f94SIra Weiny SEQ_OPTS_PUTS("dax"); 25309cb20f94SIra Weiny else 25319cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=always"); 25329cb20f94SIra Weiny } else if (test_opt2(sb, DAX_NEVER)) { 25339cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=never"); 25349cb20f94SIra Weiny } else if (test_opt2(sb, DAX_INODE)) { 25359cb20f94SIra Weiny SEQ_OPTS_PUTS("dax=inode"); 25369cb20f94SIra Weiny } 25379cb20f94SIra Weiny 25382adf6da8STheodore Ts'o ext4_show_quota_options(seq, sb); 25392adf6da8STheodore Ts'o return 0; 25402adf6da8STheodore Ts'o } 25412adf6da8STheodore Ts'o 254266acdcf4STheodore Ts'o static int ext4_show_options(struct seq_file *seq, struct dentry *root) 254366acdcf4STheodore Ts'o { 254466acdcf4STheodore Ts'o return _ext4_show_options(seq, root->d_sb, 0); 254566acdcf4STheodore Ts'o } 254666acdcf4STheodore Ts'o 2547ebd173beSTheodore Ts'o int ext4_seq_options_show(struct seq_file *seq, void *offset) 254866acdcf4STheodore Ts'o { 254966acdcf4STheodore Ts'o struct super_block *sb = seq->private; 255066acdcf4STheodore Ts'o int rc; 255166acdcf4STheodore Ts'o 2552bc98a42cSDavid Howells seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw"); 255366acdcf4STheodore Ts'o rc = _ext4_show_options(seq, sb, 1); 255466acdcf4STheodore Ts'o seq_puts(seq, "\n"); 255566acdcf4STheodore Ts'o return rc; 255666acdcf4STheodore Ts'o } 255766acdcf4STheodore Ts'o 2558617ba13bSMingming Cao static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es, 2559ac27a0ecSDave Kleikamp int read_only) 2560ac27a0ecSDave Kleikamp { 2561617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2562c89128a0SJaegeuk Kim int err = 0; 2563ac27a0ecSDave Kleikamp 2564617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) { 2565b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "revision level too high, " 2566b31e1552SEric Sandeen "forcing read-only mode"); 2567c89128a0SJaegeuk Kim err = -EROFS; 25685adaccacSyangerkun goto done; 2569ac27a0ecSDave Kleikamp } 2570ac27a0ecSDave Kleikamp if (read_only) 2571281b5995STheodore Ts'o goto done; 2572617ba13bSMingming Cao if (!(sbi->s_mount_state & EXT4_VALID_FS)) 2573b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, " 2574b31e1552SEric Sandeen "running e2fsck is recommended"); 2575c8b459f4SLukas Czerner else if (sbi->s_mount_state & EXT4_ERROR_FS) 2576b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2577b31e1552SEric Sandeen "warning: mounting fs with errors, " 2578b31e1552SEric Sandeen "running e2fsck is recommended"); 2579ed3ce80aSTao Ma else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 && 2580ac27a0ecSDave Kleikamp le16_to_cpu(es->s_mnt_count) >= 2581ac27a0ecSDave Kleikamp (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count)) 2582b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2583b31e1552SEric Sandeen "warning: maximal mount count reached, " 2584b31e1552SEric Sandeen "running e2fsck is recommended"); 2585ac27a0ecSDave Kleikamp else if (le32_to_cpu(es->s_checkinterval) && 25866a0678a7SArnd Bergmann (ext4_get_tstamp(es, s_lastcheck) + 25876a0678a7SArnd Bergmann le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds())) 2588b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 2589b31e1552SEric Sandeen "warning: checktime reached, " 2590b31e1552SEric Sandeen "running e2fsck is recommended"); 25910390131bSFrank Mayhar if (!sbi->s_journal) 2592216c34b2SMarcin Slusarz es->s_state &= cpu_to_le16(~EXT4_VALID_FS); 2593ac27a0ecSDave Kleikamp if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 2594617ba13bSMingming Cao es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 2595e8546d06SMarcin Slusarz le16_add_cpu(&es->s_mnt_count, 1); 25966a0678a7SArnd Bergmann ext4_update_tstamp(es, s_mtime); 25970390131bSFrank Mayhar if (sbi->s_journal) 2598e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 2599ac27a0ecSDave Kleikamp 2600c89128a0SJaegeuk Kim err = ext4_commit_super(sb, 1); 2601281b5995STheodore Ts'o done: 2602ac27a0ecSDave Kleikamp if (test_opt(sb, DEBUG)) 2603a9df9a49STheodore Ts'o printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, " 2604a2595b8aSTheodore Ts'o "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n", 2605ac27a0ecSDave Kleikamp sb->s_blocksize, 2606ac27a0ecSDave Kleikamp sbi->s_groups_count, 2607617ba13bSMingming Cao EXT4_BLOCKS_PER_GROUP(sb), 2608617ba13bSMingming Cao EXT4_INODES_PER_GROUP(sb), 2609a2595b8aSTheodore Ts'o sbi->s_mount_opt, sbi->s_mount_opt2); 2610ac27a0ecSDave Kleikamp 26117abc52c2SDan Magenheimer cleancache_init_fs(sb); 2612c89128a0SJaegeuk Kim return err; 2613ac27a0ecSDave Kleikamp } 2614ac27a0ecSDave Kleikamp 2615117fff10STheodore Ts'o int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup) 2616117fff10STheodore Ts'o { 2617117fff10STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 26187c990728SSuraj Jitindar Singh struct flex_groups **old_groups, **new_groups; 261937b0b6b8SDan Carpenter int size, i, j; 2620117fff10STheodore Ts'o 2621117fff10STheodore Ts'o if (!sbi->s_log_groups_per_flex) 2622117fff10STheodore Ts'o return 0; 2623117fff10STheodore Ts'o 2624117fff10STheodore Ts'o size = ext4_flex_group(sbi, ngroup - 1) + 1; 2625117fff10STheodore Ts'o if (size <= sbi->s_flex_groups_allocated) 2626117fff10STheodore Ts'o return 0; 2627117fff10STheodore Ts'o 26287c990728SSuraj Jitindar Singh new_groups = kvzalloc(roundup_pow_of_two(size * 26297c990728SSuraj Jitindar Singh sizeof(*sbi->s_flex_groups)), GFP_KERNEL); 2630117fff10STheodore Ts'o if (!new_groups) { 26317c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 26327c990728SSuraj Jitindar Singh "not enough memory for %d flex group pointers", size); 2633117fff10STheodore Ts'o return -ENOMEM; 2634117fff10STheodore Ts'o } 26357c990728SSuraj Jitindar Singh for (i = sbi->s_flex_groups_allocated; i < size; i++) { 26367c990728SSuraj Jitindar Singh new_groups[i] = kvzalloc(roundup_pow_of_two( 26377c990728SSuraj Jitindar Singh sizeof(struct flex_groups)), 26387c990728SSuraj Jitindar Singh GFP_KERNEL); 26397c990728SSuraj Jitindar Singh if (!new_groups[i]) { 264037b0b6b8SDan Carpenter for (j = sbi->s_flex_groups_allocated; j < i; j++) 264137b0b6b8SDan Carpenter kvfree(new_groups[j]); 26427c990728SSuraj Jitindar Singh kvfree(new_groups); 26437c990728SSuraj Jitindar Singh ext4_msg(sb, KERN_ERR, 26447c990728SSuraj Jitindar Singh "not enough memory for %d flex groups", size); 26457c990728SSuraj Jitindar Singh return -ENOMEM; 2646117fff10STheodore Ts'o } 26477c990728SSuraj Jitindar Singh } 26487c990728SSuraj Jitindar Singh rcu_read_lock(); 26497c990728SSuraj Jitindar Singh old_groups = rcu_dereference(sbi->s_flex_groups); 26507c990728SSuraj Jitindar Singh if (old_groups) 26517c990728SSuraj Jitindar Singh memcpy(new_groups, old_groups, 26527c990728SSuraj Jitindar Singh (sbi->s_flex_groups_allocated * 26537c990728SSuraj Jitindar Singh sizeof(struct flex_groups *))); 26547c990728SSuraj Jitindar Singh rcu_read_unlock(); 26557c990728SSuraj Jitindar Singh rcu_assign_pointer(sbi->s_flex_groups, new_groups); 26567c990728SSuraj Jitindar Singh sbi->s_flex_groups_allocated = size; 26577c990728SSuraj Jitindar Singh if (old_groups) 26587c990728SSuraj Jitindar Singh ext4_kvfree_array_rcu(old_groups); 2659117fff10STheodore Ts'o return 0; 2660117fff10STheodore Ts'o } 2661117fff10STheodore Ts'o 2662772cb7c8SJose R. Santos static int ext4_fill_flex_info(struct super_block *sb) 2663772cb7c8SJose R. Santos { 2664772cb7c8SJose R. Santos struct ext4_sb_info *sbi = EXT4_SB(sb); 2665772cb7c8SJose R. Santos struct ext4_group_desc *gdp = NULL; 26667c990728SSuraj Jitindar Singh struct flex_groups *fg; 2667772cb7c8SJose R. Santos ext4_group_t flex_group; 2668117fff10STheodore Ts'o int i, err; 2669772cb7c8SJose R. Santos 2670503358aeSTheodore Ts'o sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex; 2671d50f2ab6SXi Wang if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) { 2672772cb7c8SJose R. Santos sbi->s_log_groups_per_flex = 0; 2673772cb7c8SJose R. Santos return 1; 2674772cb7c8SJose R. Santos } 2675772cb7c8SJose R. Santos 2676117fff10STheodore Ts'o err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count); 2677117fff10STheodore Ts'o if (err) 2678772cb7c8SJose R. Santos goto failed; 2679772cb7c8SJose R. Santos 2680772cb7c8SJose R. Santos for (i = 0; i < sbi->s_groups_count; i++) { 268188b6edd1STheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 2682772cb7c8SJose R. Santos 2683772cb7c8SJose R. Santos flex_group = ext4_flex_group(sbi, i); 26847c990728SSuraj Jitindar Singh fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group); 26857c990728SSuraj Jitindar Singh atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes); 268690ba983fSTheodore Ts'o atomic64_add(ext4_free_group_clusters(sb, gdp), 26877c990728SSuraj Jitindar Singh &fg->free_clusters); 26887c990728SSuraj Jitindar Singh atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs); 2689772cb7c8SJose R. Santos } 2690772cb7c8SJose R. Santos 2691772cb7c8SJose R. Santos return 1; 2692772cb7c8SJose R. Santos failed: 2693772cb7c8SJose R. Santos return 0; 2694772cb7c8SJose R. Santos } 2695772cb7c8SJose R. Santos 2696e2b911c5SDarrick J. Wong static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group, 2697717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2698717d50e4SAndreas Dilger { 2699b47820edSDaeho Jeong int offset = offsetof(struct ext4_group_desc, bg_checksum); 2700717d50e4SAndreas Dilger __u16 crc = 0; 2701717d50e4SAndreas Dilger __le32 le_group = cpu_to_le32(block_group); 2702e2b911c5SDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 2703717d50e4SAndreas Dilger 27049aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sbi->s_sb)) { 2705feb0ab32SDarrick J. Wong /* Use new metadata_csum algorithm */ 2706feb0ab32SDarrick J. Wong __u32 csum32; 2707b47820edSDaeho Jeong __u16 dummy_csum = 0; 2708feb0ab32SDarrick J. Wong 2709feb0ab32SDarrick J. Wong csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group, 2710feb0ab32SDarrick J. Wong sizeof(le_group)); 2711b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset); 2712b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum, 2713b47820edSDaeho Jeong sizeof(dummy_csum)); 2714b47820edSDaeho Jeong offset += sizeof(dummy_csum); 2715b47820edSDaeho Jeong if (offset < sbi->s_desc_size) 2716b47820edSDaeho Jeong csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset, 2717b47820edSDaeho Jeong sbi->s_desc_size - offset); 2718feb0ab32SDarrick J. Wong 2719feb0ab32SDarrick J. Wong crc = csum32 & 0xFFFF; 2720feb0ab32SDarrick J. Wong goto out; 2721feb0ab32SDarrick J. Wong } 2722feb0ab32SDarrick J. Wong 2723feb0ab32SDarrick J. Wong /* old crc16 code */ 2724e2b911c5SDarrick J. Wong if (!ext4_has_feature_gdt_csum(sb)) 2725813d32f9SDarrick J. Wong return 0; 2726813d32f9SDarrick J. Wong 2727717d50e4SAndreas Dilger crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid)); 2728717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group)); 2729717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp, offset); 2730717d50e4SAndreas Dilger offset += sizeof(gdp->bg_checksum); /* skip checksum */ 2731717d50e4SAndreas Dilger /* for checksum of struct ext4_group_desc do the rest...*/ 2732e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 2733717d50e4SAndreas Dilger offset < le16_to_cpu(sbi->s_es->s_desc_size)) 2734717d50e4SAndreas Dilger crc = crc16(crc, (__u8 *)gdp + offset, 2735717d50e4SAndreas Dilger le16_to_cpu(sbi->s_es->s_desc_size) - 2736717d50e4SAndreas Dilger offset); 2737717d50e4SAndreas Dilger 2738feb0ab32SDarrick J. Wong out: 2739717d50e4SAndreas Dilger return cpu_to_le16(crc); 2740717d50e4SAndreas Dilger } 2741717d50e4SAndreas Dilger 2742feb0ab32SDarrick J. Wong int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group, 2743717d50e4SAndreas Dilger struct ext4_group_desc *gdp) 2744717d50e4SAndreas Dilger { 2745feb0ab32SDarrick J. Wong if (ext4_has_group_desc_csum(sb) && 2746e2b911c5SDarrick J. Wong (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp))) 2747717d50e4SAndreas Dilger return 0; 2748717d50e4SAndreas Dilger 2749717d50e4SAndreas Dilger return 1; 2750717d50e4SAndreas Dilger } 2751717d50e4SAndreas Dilger 2752feb0ab32SDarrick J. Wong void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group, 2753feb0ab32SDarrick J. Wong struct ext4_group_desc *gdp) 2754feb0ab32SDarrick J. Wong { 2755feb0ab32SDarrick J. Wong if (!ext4_has_group_desc_csum(sb)) 2756feb0ab32SDarrick J. Wong return; 2757e2b911c5SDarrick J. Wong gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp); 2758feb0ab32SDarrick J. Wong } 2759feb0ab32SDarrick J. Wong 2760ac27a0ecSDave Kleikamp /* Called at mount-time, super-block is locked */ 2761bfff6873SLukas Czerner static int ext4_check_descriptors(struct super_block *sb, 2762829fa70dSTheodore Ts'o ext4_fsblk_t sb_block, 2763bfff6873SLukas Czerner ext4_group_t *first_not_zeroed) 2764ac27a0ecSDave Kleikamp { 2765617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 2766617ba13bSMingming Cao ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 2767617ba13bSMingming Cao ext4_fsblk_t last_block; 276844de022cSTheodore Ts'o ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0); 2769bd81d8eeSLaurent Vivier ext4_fsblk_t block_bitmap; 2770bd81d8eeSLaurent Vivier ext4_fsblk_t inode_bitmap; 2771bd81d8eeSLaurent Vivier ext4_fsblk_t inode_table; 2772ce421581SJose R. Santos int flexbg_flag = 0; 2773bfff6873SLukas Czerner ext4_group_t i, grp = sbi->s_groups_count; 2774ac27a0ecSDave Kleikamp 2775e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 2776ce421581SJose R. Santos flexbg_flag = 1; 2777ce421581SJose R. Santos 2778617ba13bSMingming Cao ext4_debug("Checking group descriptors"); 2779ac27a0ecSDave Kleikamp 2780197cd65aSAkinobu Mita for (i = 0; i < sbi->s_groups_count; i++) { 2781197cd65aSAkinobu Mita struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL); 2782197cd65aSAkinobu Mita 2783ce421581SJose R. Santos if (i == sbi->s_groups_count - 1 || flexbg_flag) 2784bd81d8eeSLaurent Vivier last_block = ext4_blocks_count(sbi->s_es) - 1; 2785ac27a0ecSDave Kleikamp else 2786ac27a0ecSDave Kleikamp last_block = first_block + 2787617ba13bSMingming Cao (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2788ac27a0ecSDave Kleikamp 2789bfff6873SLukas Czerner if ((grp == sbi->s_groups_count) && 2790bfff6873SLukas Czerner !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 2791bfff6873SLukas Czerner grp = i; 2792bfff6873SLukas Czerner 27938fadc143SAlexandre Ratchov block_bitmap = ext4_block_bitmap(sb, gdp); 2794829fa70dSTheodore Ts'o if (block_bitmap == sb_block) { 2795829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2796829fa70dSTheodore Ts'o "Block bitmap for group %u overlaps " 2797829fa70dSTheodore Ts'o "superblock", i); 279818db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 279918db4b4eSTheodore Ts'o return 0; 2800829fa70dSTheodore Ts'o } 280177260807STheodore Ts'o if (block_bitmap >= sb_block + 1 && 280277260807STheodore Ts'o block_bitmap <= last_bg_block) { 280377260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 280477260807STheodore Ts'o "Block bitmap for group %u overlaps " 280577260807STheodore Ts'o "block group descriptors", i); 280677260807STheodore Ts'o if (!sb_rdonly(sb)) 280777260807STheodore Ts'o return 0; 280877260807STheodore Ts'o } 28092b2d6d01STheodore Ts'o if (block_bitmap < first_block || block_bitmap > last_block) { 2810b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2811a9df9a49STheodore Ts'o "Block bitmap for group %u not in group " 2812b31e1552SEric Sandeen "(block %llu)!", i, block_bitmap); 2813ac27a0ecSDave Kleikamp return 0; 2814ac27a0ecSDave Kleikamp } 28158fadc143SAlexandre Ratchov inode_bitmap = ext4_inode_bitmap(sb, gdp); 2816829fa70dSTheodore Ts'o if (inode_bitmap == sb_block) { 2817829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2818829fa70dSTheodore Ts'o "Inode bitmap for group %u overlaps " 2819829fa70dSTheodore Ts'o "superblock", i); 282018db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 282118db4b4eSTheodore Ts'o return 0; 2822829fa70dSTheodore Ts'o } 282377260807STheodore Ts'o if (inode_bitmap >= sb_block + 1 && 282477260807STheodore Ts'o inode_bitmap <= last_bg_block) { 282577260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 282677260807STheodore Ts'o "Inode bitmap for group %u overlaps " 282777260807STheodore Ts'o "block group descriptors", i); 282877260807STheodore Ts'o if (!sb_rdonly(sb)) 282977260807STheodore Ts'o return 0; 283077260807STheodore Ts'o } 28312b2d6d01STheodore Ts'o if (inode_bitmap < first_block || inode_bitmap > last_block) { 2832b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2833a9df9a49STheodore Ts'o "Inode bitmap for group %u not in group " 2834b31e1552SEric Sandeen "(block %llu)!", i, inode_bitmap); 2835ac27a0ecSDave Kleikamp return 0; 2836ac27a0ecSDave Kleikamp } 28378fadc143SAlexandre Ratchov inode_table = ext4_inode_table(sb, gdp); 2838829fa70dSTheodore Ts'o if (inode_table == sb_block) { 2839829fa70dSTheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2840829fa70dSTheodore Ts'o "Inode table for group %u overlaps " 2841829fa70dSTheodore Ts'o "superblock", i); 284218db4b4eSTheodore Ts'o if (!sb_rdonly(sb)) 284318db4b4eSTheodore Ts'o return 0; 2844829fa70dSTheodore Ts'o } 284577260807STheodore Ts'o if (inode_table >= sb_block + 1 && 284677260807STheodore Ts'o inode_table <= last_bg_block) { 284777260807STheodore Ts'o ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 284877260807STheodore Ts'o "Inode table for group %u overlaps " 284977260807STheodore Ts'o "block group descriptors", i); 285077260807STheodore Ts'o if (!sb_rdonly(sb)) 285177260807STheodore Ts'o return 0; 285277260807STheodore Ts'o } 2853bd81d8eeSLaurent Vivier if (inode_table < first_block || 28542b2d6d01STheodore Ts'o inode_table + sbi->s_itb_per_group - 1 > last_block) { 2855b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2856a9df9a49STheodore Ts'o "Inode table for group %u not in group " 2857b31e1552SEric Sandeen "(block %llu)!", i, inode_table); 2858ac27a0ecSDave Kleikamp return 0; 2859ac27a0ecSDave Kleikamp } 2860955ce5f5SAneesh Kumar K.V ext4_lock_group(sb, i); 2861feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, i, gdp)) { 2862b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2863b31e1552SEric Sandeen "Checksum for group %u failed (%u!=%u)", 2864e2b911c5SDarrick J. Wong i, le16_to_cpu(ext4_group_desc_csum(sb, i, 2865fd2d4291SAvantika Mathur gdp)), le16_to_cpu(gdp->bg_checksum)); 2866bc98a42cSDavid Howells if (!sb_rdonly(sb)) { 2867955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2868717d50e4SAndreas Dilger return 0; 2869717d50e4SAndreas Dilger } 28707ee1ec4cSLi Zefan } 2871955ce5f5SAneesh Kumar K.V ext4_unlock_group(sb, i); 2872ce421581SJose R. Santos if (!flexbg_flag) 2873617ba13bSMingming Cao first_block += EXT4_BLOCKS_PER_GROUP(sb); 2874ac27a0ecSDave Kleikamp } 2875bfff6873SLukas Czerner if (NULL != first_not_zeroed) 2876bfff6873SLukas Czerner *first_not_zeroed = grp; 2877ac27a0ecSDave Kleikamp return 1; 2878ac27a0ecSDave Kleikamp } 2879ac27a0ecSDave Kleikamp 2880617ba13bSMingming Cao /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at 2881ac27a0ecSDave Kleikamp * the superblock) which were deleted from all directories, but held open by 2882ac27a0ecSDave Kleikamp * a process at the time of a crash. We walk the list and try to delete these 2883ac27a0ecSDave Kleikamp * inodes at recovery time (only with a read-write filesystem). 2884ac27a0ecSDave Kleikamp * 2885ac27a0ecSDave Kleikamp * In order to keep the orphan inode chain consistent during traversal (in 2886ac27a0ecSDave Kleikamp * case of crash during recovery), we link each inode into the superblock 2887ac27a0ecSDave Kleikamp * orphan list_head and handle it the same way as an inode deletion during 2888ac27a0ecSDave Kleikamp * normal operation (which journals the operations for us). 2889ac27a0ecSDave Kleikamp * 2890ac27a0ecSDave Kleikamp * We only do an iget() and an iput() on each inode, which is very safe if we 2891ac27a0ecSDave Kleikamp * accidentally point at an in-use or already deleted inode. The worst that 2892ac27a0ecSDave Kleikamp * can happen in this case is that we get a "bit already cleared" message from 2893617ba13bSMingming Cao * ext4_free_inode(). The only reason we would point at a wrong inode is if 2894ac27a0ecSDave Kleikamp * e2fsck was run on this filesystem, and it must have already done the orphan 2895ac27a0ecSDave Kleikamp * inode cleanup for us, so we can safely abort without any further action. 2896ac27a0ecSDave Kleikamp */ 2897617ba13bSMingming Cao static void ext4_orphan_cleanup(struct super_block *sb, 2898617ba13bSMingming Cao struct ext4_super_block *es) 2899ac27a0ecSDave Kleikamp { 2900ac27a0ecSDave Kleikamp unsigned int s_flags = sb->s_flags; 29012c98eb5eSTheodore Ts'o int ret, nr_orphans = 0, nr_truncates = 0; 2902ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 290395f1fda4Szhangyi (F) int quota_update = 0; 2904ac27a0ecSDave Kleikamp int i; 2905ac27a0ecSDave Kleikamp #endif 2906ac27a0ecSDave Kleikamp if (!es->s_last_orphan) { 2907ac27a0ecSDave Kleikamp jbd_debug(4, "no orphan inodes to clean up\n"); 2908ac27a0ecSDave Kleikamp return; 2909ac27a0ecSDave Kleikamp } 2910ac27a0ecSDave Kleikamp 2911a8f48a95SEric Sandeen if (bdev_read_only(sb->s_bdev)) { 2912b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 2913b31e1552SEric Sandeen "unavailable, skipping orphan cleanup"); 2914a8f48a95SEric Sandeen return; 2915a8f48a95SEric Sandeen } 2916a8f48a95SEric Sandeen 2917d39195c3SAmir Goldstein /* Check if feature set would not allow a r/w mount */ 2918d39195c3SAmir Goldstein if (!ext4_feature_set_ok(sb, 0)) { 2919d39195c3SAmir Goldstein ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to " 2920d39195c3SAmir Goldstein "unknown ROCOMPAT features"); 2921d39195c3SAmir Goldstein return; 2922d39195c3SAmir Goldstein } 2923d39195c3SAmir Goldstein 2924617ba13bSMingming Cao if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2925c25f9bc6SEric Sandeen /* don't clear list on RO mount w/ errors */ 29261751e8a6SLinus Torvalds if (es->s_last_orphan && !(s_flags & SB_RDONLY)) { 292784474976SDmitry Monakhov ext4_msg(sb, KERN_INFO, "Errors on filesystem, " 2928ac27a0ecSDave Kleikamp "clearing orphan list.\n"); 2929ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2930c25f9bc6SEric Sandeen } 2931ac27a0ecSDave Kleikamp jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2932ac27a0ecSDave Kleikamp return; 2933ac27a0ecSDave Kleikamp } 2934ac27a0ecSDave Kleikamp 29351751e8a6SLinus Torvalds if (s_flags & SB_RDONLY) { 2936b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs"); 29371751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 2938ac27a0ecSDave Kleikamp } 2939ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 2940ac27a0ecSDave Kleikamp /* Needed for iput() to work correctly and not trash data */ 29411751e8a6SLinus Torvalds sb->s_flags |= SB_ACTIVE; 294295f1fda4Szhangyi (F) 294395f1fda4Szhangyi (F) /* 294495f1fda4Szhangyi (F) * Turn on quotas which were not enabled for read-only mounts if 294595f1fda4Szhangyi (F) * filesystem has quota feature, so that they are updated correctly. 294695f1fda4Szhangyi (F) */ 29471751e8a6SLinus Torvalds if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) { 294895f1fda4Szhangyi (F) int ret = ext4_enable_quotas(sb); 294995f1fda4Szhangyi (F) 295095f1fda4Szhangyi (F) if (!ret) 295195f1fda4Szhangyi (F) quota_update = 1; 295295f1fda4Szhangyi (F) else 295395f1fda4Szhangyi (F) ext4_msg(sb, KERN_ERR, 295495f1fda4Szhangyi (F) "Cannot turn on quotas: error %d", ret); 295595f1fda4Szhangyi (F) } 295695f1fda4Szhangyi (F) 295795f1fda4Szhangyi (F) /* Turn on journaled quotas used for old sytle */ 2958a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 2959617ba13bSMingming Cao if (EXT4_SB(sb)->s_qf_names[i]) { 2960617ba13bSMingming Cao int ret = ext4_quota_on_mount(sb, i); 296195f1fda4Szhangyi (F) 296295f1fda4Szhangyi (F) if (!ret) 296395f1fda4Szhangyi (F) quota_update = 1; 296495f1fda4Szhangyi (F) else 2965b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 2966b31e1552SEric Sandeen "Cannot turn on journaled " 296795f1fda4Szhangyi (F) "quota: type %d: error %d", i, ret); 2968ac27a0ecSDave Kleikamp } 2969ac27a0ecSDave Kleikamp } 2970ac27a0ecSDave Kleikamp #endif 2971ac27a0ecSDave Kleikamp 2972ac27a0ecSDave Kleikamp while (es->s_last_orphan) { 2973ac27a0ecSDave Kleikamp struct inode *inode; 2974ac27a0ecSDave Kleikamp 2975c65d5c6cSVegard Nossum /* 2976c65d5c6cSVegard Nossum * We may have encountered an error during cleanup; if 2977c65d5c6cSVegard Nossum * so, skip the rest. 2978c65d5c6cSVegard Nossum */ 2979c65d5c6cSVegard Nossum if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) { 2980c65d5c6cSVegard Nossum jbd_debug(1, "Skipping orphan recovery on fs with errors.\n"); 2981c65d5c6cSVegard Nossum es->s_last_orphan = 0; 2982c65d5c6cSVegard Nossum break; 2983c65d5c6cSVegard Nossum } 2984c65d5c6cSVegard Nossum 298597bd42b9SJosef Bacik inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)); 298697bd42b9SJosef Bacik if (IS_ERR(inode)) { 2987ac27a0ecSDave Kleikamp es->s_last_orphan = 0; 2988ac27a0ecSDave Kleikamp break; 2989ac27a0ecSDave Kleikamp } 2990ac27a0ecSDave Kleikamp 2991617ba13bSMingming Cao list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); 2992871a2931SChristoph Hellwig dquot_initialize(inode); 2993ac27a0ecSDave Kleikamp if (inode->i_nlink) { 2994566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 2995b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 2996b31e1552SEric Sandeen "%s: truncating inode %lu to %lld bytes", 299746e665e9SHarvey Harrison __func__, inode->i_ino, inode->i_size); 2998e5f8eab8STheodore Ts'o jbd_debug(2, "truncating inode %lu to %lld bytes\n", 2999ac27a0ecSDave Kleikamp inode->i_ino, inode->i_size); 30005955102cSAl Viro inode_lock(inode); 300155f252c9SLukas Czerner truncate_inode_pages(inode->i_mapping, inode->i_size); 30022c98eb5eSTheodore Ts'o ret = ext4_truncate(inode); 30032c98eb5eSTheodore Ts'o if (ret) 30042c98eb5eSTheodore Ts'o ext4_std_error(inode->i_sb, ret); 30055955102cSAl Viro inode_unlock(inode); 3006ac27a0ecSDave Kleikamp nr_truncates++; 3007ac27a0ecSDave Kleikamp } else { 3008566370a2SPaul Taysom if (test_opt(sb, DEBUG)) 3009b31e1552SEric Sandeen ext4_msg(sb, KERN_DEBUG, 3010b31e1552SEric Sandeen "%s: deleting unreferenced inode %lu", 301146e665e9SHarvey Harrison __func__, inode->i_ino); 3012ac27a0ecSDave Kleikamp jbd_debug(2, "deleting unreferenced inode %lu\n", 3013ac27a0ecSDave Kleikamp inode->i_ino); 3014ac27a0ecSDave Kleikamp nr_orphans++; 3015ac27a0ecSDave Kleikamp } 3016ac27a0ecSDave Kleikamp iput(inode); /* The delete magic happens here! */ 3017ac27a0ecSDave Kleikamp } 3018ac27a0ecSDave Kleikamp 3019ac27a0ecSDave Kleikamp #define PLURAL(x) (x), ((x) == 1) ? "" : "s" 3020ac27a0ecSDave Kleikamp 3021ac27a0ecSDave Kleikamp if (nr_orphans) 3022b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted", 3023b31e1552SEric Sandeen PLURAL(nr_orphans)); 3024ac27a0ecSDave Kleikamp if (nr_truncates) 3025b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up", 3026b31e1552SEric Sandeen PLURAL(nr_truncates)); 3027ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 302895f1fda4Szhangyi (F) /* Turn off quotas if they were enabled for orphan cleanup */ 302995f1fda4Szhangyi (F) if (quota_update) { 3030a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 303195f1fda4Szhangyi (F) if (sb_dqopt(sb)->files[i]) 3032287a8095SChristoph Hellwig dquot_quota_off(sb, i); 3033ac27a0ecSDave Kleikamp } 303495f1fda4Szhangyi (F) } 3035ac27a0ecSDave Kleikamp #endif 30361751e8a6SLinus Torvalds sb->s_flags = s_flags; /* Restore SB_RDONLY status */ 3037ac27a0ecSDave Kleikamp } 30380b8e58a1SAndreas Dilger 3039cd2291a4SEric Sandeen /* 3040cd2291a4SEric Sandeen * Maximal extent format file size. 3041cd2291a4SEric Sandeen * Resulting logical blkno at s_maxbytes must fit in our on-disk 3042cd2291a4SEric Sandeen * extent format containers, within a sector_t, and within i_blocks 3043cd2291a4SEric Sandeen * in the vfs. ext4 inode has 48 bits of i_block in fsblock units, 3044cd2291a4SEric Sandeen * so that won't be a limiting factor. 3045cd2291a4SEric Sandeen * 3046f17722f9SLukas Czerner * However there is other limiting factor. We do store extents in the form 3047f17722f9SLukas Czerner * of starting block and length, hence the resulting length of the extent 3048f17722f9SLukas Czerner * covering maximum file size must fit into on-disk format containers as 3049f17722f9SLukas Czerner * well. Given that length is always by 1 unit bigger than max unit (because 3050f17722f9SLukas Czerner * we count 0 as well) we have to lower the s_maxbytes by one fs block. 3051f17722f9SLukas Czerner * 3052cd2291a4SEric Sandeen * Note, this does *not* consider any metadata overhead for vfs i_blocks. 3053cd2291a4SEric Sandeen */ 3054f287a1a5STheodore Ts'o static loff_t ext4_max_size(int blkbits, int has_huge_files) 3055cd2291a4SEric Sandeen { 3056cd2291a4SEric Sandeen loff_t res; 3057cd2291a4SEric Sandeen loff_t upper_limit = MAX_LFS_FILESIZE; 3058cd2291a4SEric Sandeen 305972deb455SChristoph Hellwig BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64)); 306072deb455SChristoph Hellwig 306172deb455SChristoph Hellwig if (!has_huge_files) { 3062cd2291a4SEric Sandeen upper_limit = (1LL << 32) - 1; 3063cd2291a4SEric Sandeen 3064cd2291a4SEric Sandeen /* total blocks in file system block size */ 3065cd2291a4SEric Sandeen upper_limit >>= (blkbits - 9); 3066cd2291a4SEric Sandeen upper_limit <<= blkbits; 3067cd2291a4SEric Sandeen } 3068cd2291a4SEric Sandeen 3069f17722f9SLukas Czerner /* 3070f17722f9SLukas Czerner * 32-bit extent-start container, ee_block. We lower the maxbytes 3071f17722f9SLukas Czerner * by one fs block, so ee_len can cover the extent of maximum file 3072f17722f9SLukas Czerner * size 3073f17722f9SLukas Czerner */ 3074f17722f9SLukas Czerner res = (1LL << 32) - 1; 3075cd2291a4SEric Sandeen res <<= blkbits; 3076cd2291a4SEric Sandeen 3077cd2291a4SEric Sandeen /* Sanity check against vm- & vfs- imposed limits */ 3078cd2291a4SEric Sandeen if (res > upper_limit) 3079cd2291a4SEric Sandeen res = upper_limit; 3080cd2291a4SEric Sandeen 3081cd2291a4SEric Sandeen return res; 3082cd2291a4SEric Sandeen } 3083ac27a0ecSDave Kleikamp 3084ac27a0ecSDave Kleikamp /* 3085cd2291a4SEric Sandeen * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect 30860fc1b451SAneesh Kumar K.V * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks. 30870fc1b451SAneesh Kumar K.V * We need to be 1 filesystem block less than the 2^48 sector limit. 3088ac27a0ecSDave Kleikamp */ 3089f287a1a5STheodore Ts'o static loff_t ext4_max_bitmap_size(int bits, int has_huge_files) 3090ac27a0ecSDave Kleikamp { 3091617ba13bSMingming Cao loff_t res = EXT4_NDIR_BLOCKS; 30920fc1b451SAneesh Kumar K.V int meta_blocks; 30930fc1b451SAneesh Kumar K.V loff_t upper_limit; 30940b8e58a1SAndreas Dilger /* This is calculated to be the largest file size for a dense, block 30950b8e58a1SAndreas Dilger * mapped file such that the file's total number of 512-byte sectors, 30960b8e58a1SAndreas Dilger * including data and all indirect blocks, does not exceed (2^48 - 1). 30970b8e58a1SAndreas Dilger * 30980b8e58a1SAndreas Dilger * __u32 i_blocks_lo and _u16 i_blocks_high represent the total 30990b8e58a1SAndreas Dilger * number of 512-byte sectors of the file. 31000fc1b451SAneesh Kumar K.V */ 31010fc1b451SAneesh Kumar K.V 310272deb455SChristoph Hellwig if (!has_huge_files) { 31030fc1b451SAneesh Kumar K.V /* 310472deb455SChristoph Hellwig * !has_huge_files or implies that the inode i_block field 310572deb455SChristoph Hellwig * represents total file blocks in 2^32 512-byte sectors == 310672deb455SChristoph Hellwig * size of vfs inode i_blocks * 8 31070fc1b451SAneesh Kumar K.V */ 31080fc1b451SAneesh Kumar K.V upper_limit = (1LL << 32) - 1; 31090fc1b451SAneesh Kumar K.V 31100fc1b451SAneesh Kumar K.V /* total blocks in file system block size */ 31110fc1b451SAneesh Kumar K.V upper_limit >>= (bits - 9); 31120fc1b451SAneesh Kumar K.V 31130fc1b451SAneesh Kumar K.V } else { 31148180a562SAneesh Kumar K.V /* 31158180a562SAneesh Kumar K.V * We use 48 bit ext4_inode i_blocks 31168180a562SAneesh Kumar K.V * With EXT4_HUGE_FILE_FL set the i_blocks 31178180a562SAneesh Kumar K.V * represent total number of blocks in 31188180a562SAneesh Kumar K.V * file system block size 31198180a562SAneesh Kumar K.V */ 31200fc1b451SAneesh Kumar K.V upper_limit = (1LL << 48) - 1; 31210fc1b451SAneesh Kumar K.V 31220fc1b451SAneesh Kumar K.V } 31230fc1b451SAneesh Kumar K.V 31240fc1b451SAneesh Kumar K.V /* indirect blocks */ 31250fc1b451SAneesh Kumar K.V meta_blocks = 1; 31260fc1b451SAneesh Kumar K.V /* double indirect blocks */ 31270fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)); 31280fc1b451SAneesh Kumar K.V /* tripple indirect blocks */ 31290fc1b451SAneesh Kumar K.V meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2))); 31300fc1b451SAneesh Kumar K.V 31310fc1b451SAneesh Kumar K.V upper_limit -= meta_blocks; 31320fc1b451SAneesh Kumar K.V upper_limit <<= bits; 3133ac27a0ecSDave Kleikamp 3134ac27a0ecSDave Kleikamp res += 1LL << (bits-2); 3135ac27a0ecSDave Kleikamp res += 1LL << (2*(bits-2)); 3136ac27a0ecSDave Kleikamp res += 1LL << (3*(bits-2)); 3137ac27a0ecSDave Kleikamp res <<= bits; 3138ac27a0ecSDave Kleikamp if (res > upper_limit) 3139ac27a0ecSDave Kleikamp res = upper_limit; 31400fc1b451SAneesh Kumar K.V 31410fc1b451SAneesh Kumar K.V if (res > MAX_LFS_FILESIZE) 31420fc1b451SAneesh Kumar K.V res = MAX_LFS_FILESIZE; 31430fc1b451SAneesh Kumar K.V 3144ac27a0ecSDave Kleikamp return res; 3145ac27a0ecSDave Kleikamp } 3146ac27a0ecSDave Kleikamp 3147617ba13bSMingming Cao static ext4_fsblk_t descriptor_loc(struct super_block *sb, 314870bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block, int nr) 3149ac27a0ecSDave Kleikamp { 3150617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 3151fd2d4291SAvantika Mathur ext4_group_t bg, first_meta_bg; 3152ac27a0ecSDave Kleikamp int has_super = 0; 3153ac27a0ecSDave Kleikamp 3154ac27a0ecSDave Kleikamp first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); 3155ac27a0ecSDave Kleikamp 3156e2b911c5SDarrick J. Wong if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg) 315770bbb3e0SAndrew Morton return logical_sb_block + nr + 1; 3158ac27a0ecSDave Kleikamp bg = sbi->s_desc_per_block * nr; 3159617ba13bSMingming Cao if (ext4_bg_has_super(sb, bg)) 3160ac27a0ecSDave Kleikamp has_super = 1; 31610b8e58a1SAndreas Dilger 3162bd63f6b0SDarrick J. Wong /* 3163bd63f6b0SDarrick J. Wong * If we have a meta_bg fs with 1k blocks, group 0's GDT is at 3164bd63f6b0SDarrick J. Wong * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled 3165bd63f6b0SDarrick J. Wong * on modern mke2fs or blksize > 1k on older mke2fs) then we must 3166bd63f6b0SDarrick J. Wong * compensate. 3167bd63f6b0SDarrick J. Wong */ 3168bd63f6b0SDarrick J. Wong if (sb->s_blocksize == 1024 && nr == 0 && 316949598e04SJun Piao le32_to_cpu(sbi->s_es->s_first_data_block) == 0) 3170bd63f6b0SDarrick J. Wong has_super++; 3171bd63f6b0SDarrick J. Wong 3172617ba13bSMingming Cao return (has_super + ext4_group_first_block_no(sb, bg)); 3173ac27a0ecSDave Kleikamp } 3174ac27a0ecSDave Kleikamp 3175c9de560dSAlex Tomas /** 3176c9de560dSAlex Tomas * ext4_get_stripe_size: Get the stripe size. 3177c9de560dSAlex Tomas * @sbi: In memory super block info 3178c9de560dSAlex Tomas * 3179c9de560dSAlex Tomas * If we have specified it via mount option, then 3180c9de560dSAlex Tomas * use the mount option value. If the value specified at mount time is 3181c9de560dSAlex Tomas * greater than the blocks per group use the super block value. 3182c9de560dSAlex Tomas * If the super block value is greater than blocks per group return 0. 3183c9de560dSAlex Tomas * Allocator needs it be less than blocks per group. 3184c9de560dSAlex Tomas * 3185c9de560dSAlex Tomas */ 3186c9de560dSAlex Tomas static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi) 3187c9de560dSAlex Tomas { 3188c9de560dSAlex Tomas unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride); 3189c9de560dSAlex Tomas unsigned long stripe_width = 3190c9de560dSAlex Tomas le32_to_cpu(sbi->s_es->s_raid_stripe_width); 31913eb08658SDan Ehrenberg int ret; 3192c9de560dSAlex Tomas 3193c9de560dSAlex Tomas if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group) 31943eb08658SDan Ehrenberg ret = sbi->s_stripe; 31955469d7c3SJan Kara else if (stripe_width && stripe_width <= sbi->s_blocks_per_group) 31963eb08658SDan Ehrenberg ret = stripe_width; 31975469d7c3SJan Kara else if (stride && stride <= sbi->s_blocks_per_group) 31983eb08658SDan Ehrenberg ret = stride; 31993eb08658SDan Ehrenberg else 32003eb08658SDan Ehrenberg ret = 0; 3201c9de560dSAlex Tomas 32023eb08658SDan Ehrenberg /* 32033eb08658SDan Ehrenberg * If the stripe width is 1, this makes no sense and 32043eb08658SDan Ehrenberg * we set it to 0 to turn off stripe handling code. 32053eb08658SDan Ehrenberg */ 32063eb08658SDan Ehrenberg if (ret <= 1) 32073eb08658SDan Ehrenberg ret = 0; 3208c9de560dSAlex Tomas 32093eb08658SDan Ehrenberg return ret; 3210c9de560dSAlex Tomas } 3211ac27a0ecSDave Kleikamp 3212a13fb1a4SEric Sandeen /* 3213a13fb1a4SEric Sandeen * Check whether this filesystem can be mounted based on 3214a13fb1a4SEric Sandeen * the features present and the RDONLY/RDWR mount requested. 3215a13fb1a4SEric Sandeen * Returns 1 if this filesystem can be mounted as requested, 3216a13fb1a4SEric Sandeen * 0 if it cannot be. 3217a13fb1a4SEric Sandeen */ 3218a13fb1a4SEric Sandeen static int ext4_feature_set_ok(struct super_block *sb, int readonly) 3219a13fb1a4SEric Sandeen { 3220e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_incompat_features(sb)) { 3221a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, 3222a13fb1a4SEric Sandeen "Couldn't mount because of " 3223a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3224a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) & 3225a13fb1a4SEric Sandeen ~EXT4_FEATURE_INCOMPAT_SUPP)); 3226a13fb1a4SEric Sandeen return 0; 3227a13fb1a4SEric Sandeen } 3228a13fb1a4SEric Sandeen 3229c83ad55eSGabriel Krisman Bertazi #ifndef CONFIG_UNICODE 3230c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb)) { 3231c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 3232c83ad55eSGabriel Krisman Bertazi "Filesystem with casefold feature cannot be " 3233c83ad55eSGabriel Krisman Bertazi "mounted without CONFIG_UNICODE"); 3234c83ad55eSGabriel Krisman Bertazi return 0; 3235c83ad55eSGabriel Krisman Bertazi } 3236c83ad55eSGabriel Krisman Bertazi #endif 3237c83ad55eSGabriel Krisman Bertazi 3238a13fb1a4SEric Sandeen if (readonly) 3239a13fb1a4SEric Sandeen return 1; 3240a13fb1a4SEric Sandeen 3241e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb)) { 32422cb5cc8bSDarrick J. Wong ext4_msg(sb, KERN_INFO, "filesystem is read-only"); 32431751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 32442cb5cc8bSDarrick J. Wong return 1; 32452cb5cc8bSDarrick J. Wong } 32462cb5cc8bSDarrick J. Wong 3247a13fb1a4SEric Sandeen /* Check that feature set is OK for a read-write mount */ 3248e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext4_ro_compat_features(sb)) { 3249a13fb1a4SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of " 3250a13fb1a4SEric Sandeen "unsupported optional features (%x)", 3251a13fb1a4SEric Sandeen (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) & 3252a13fb1a4SEric Sandeen ~EXT4_FEATURE_RO_COMPAT_SUPP)); 3253a13fb1a4SEric Sandeen return 0; 3254a13fb1a4SEric Sandeen } 3255e2b911c5SDarrick J. Wong if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) { 3256bab08ab9STheodore Ts'o ext4_msg(sb, KERN_ERR, 3257bab08ab9STheodore Ts'o "Can't support bigalloc feature without " 3258bab08ab9STheodore Ts'o "extents feature\n"); 3259bab08ab9STheodore Ts'o return 0; 3260bab08ab9STheodore Ts'o } 32617c319d32SAditya Kali 32629db176bcSJan Kara #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2) 3263d65d87a0STheodore Ts'o if (!readonly && (ext4_has_feature_quota(sb) || 3264d65d87a0STheodore Ts'o ext4_has_feature_project(sb))) { 32657c319d32SAditya Kali ext4_msg(sb, KERN_ERR, 3266d65d87a0STheodore Ts'o "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2"); 3267689c958cSLi Xi return 0; 3268689c958cSLi Xi } 32697c319d32SAditya Kali #endif /* CONFIG_QUOTA */ 3270a13fb1a4SEric Sandeen return 1; 3271a13fb1a4SEric Sandeen } 3272a13fb1a4SEric Sandeen 327366e61a9eSTheodore Ts'o /* 327466e61a9eSTheodore Ts'o * This function is called once a day if we have errors logged 327566e61a9eSTheodore Ts'o * on the file system 327666e61a9eSTheodore Ts'o */ 3277235699a8SKees Cook static void print_daily_error_info(struct timer_list *t) 327866e61a9eSTheodore Ts'o { 3279235699a8SKees Cook struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report); 3280235699a8SKees Cook struct super_block *sb = sbi->s_sb; 3281235699a8SKees Cook struct ext4_super_block *es = sbi->s_es; 328266e61a9eSTheodore Ts'o 328366e61a9eSTheodore Ts'o if (es->s_error_count) 3284ae0f78deSTheodore Ts'o /* fsck newer than v1.41.13 is needed to clean this condition. */ 3285ae0f78deSTheodore Ts'o ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u", 328666e61a9eSTheodore Ts'o le32_to_cpu(es->s_error_count)); 328766e61a9eSTheodore Ts'o if (es->s_first_error_time) { 32886a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d", 32896a0678a7SArnd Bergmann sb->s_id, 32906a0678a7SArnd Bergmann ext4_get_tstamp(es, s_first_error_time), 329166e61a9eSTheodore Ts'o (int) sizeof(es->s_first_error_func), 329266e61a9eSTheodore Ts'o es->s_first_error_func, 329366e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_line)); 329466e61a9eSTheodore Ts'o if (es->s_first_error_ino) 3295651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 329666e61a9eSTheodore Ts'o le32_to_cpu(es->s_first_error_ino)); 329766e61a9eSTheodore Ts'o if (es->s_first_error_block) 3298651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 329966e61a9eSTheodore Ts'o le64_to_cpu(es->s_first_error_block)); 3300651e1c3bSJoe Perches printk(KERN_CONT "\n"); 330166e61a9eSTheodore Ts'o } 330266e61a9eSTheodore Ts'o if (es->s_last_error_time) { 33036a0678a7SArnd Bergmann printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d", 33046a0678a7SArnd Bergmann sb->s_id, 33056a0678a7SArnd Bergmann ext4_get_tstamp(es, s_last_error_time), 330666e61a9eSTheodore Ts'o (int) sizeof(es->s_last_error_func), 330766e61a9eSTheodore Ts'o es->s_last_error_func, 330866e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_line)); 330966e61a9eSTheodore Ts'o if (es->s_last_error_ino) 3310651e1c3bSJoe Perches printk(KERN_CONT ": inode %u", 331166e61a9eSTheodore Ts'o le32_to_cpu(es->s_last_error_ino)); 331266e61a9eSTheodore Ts'o if (es->s_last_error_block) 3313651e1c3bSJoe Perches printk(KERN_CONT ": block %llu", (unsigned long long) 331466e61a9eSTheodore Ts'o le64_to_cpu(es->s_last_error_block)); 3315651e1c3bSJoe Perches printk(KERN_CONT "\n"); 331666e61a9eSTheodore Ts'o } 331766e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */ 331866e61a9eSTheodore Ts'o } 331966e61a9eSTheodore Ts'o 3320bfff6873SLukas Czerner /* Find next suitable group and run ext4_init_inode_table */ 3321bfff6873SLukas Czerner static int ext4_run_li_request(struct ext4_li_request *elr) 3322bfff6873SLukas Czerner { 3323bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 33243d392b26STheodore Ts'o struct super_block *sb = elr->lr_super; 33253d392b26STheodore Ts'o ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count; 33263d392b26STheodore Ts'o ext4_group_t group = elr->lr_next_group; 3327bfff6873SLukas Czerner unsigned long timeout = 0; 33283d392b26STheodore Ts'o unsigned int prefetch_ios = 0; 3329bfff6873SLukas Czerner int ret = 0; 3330bfff6873SLukas Czerner 33313d392b26STheodore Ts'o if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) { 33323d392b26STheodore Ts'o elr->lr_next_group = ext4_mb_prefetch(sb, group, 33333d392b26STheodore Ts'o EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios); 33343d392b26STheodore Ts'o if (prefetch_ios) 33353d392b26STheodore Ts'o ext4_mb_prefetch_fini(sb, elr->lr_next_group, 33363d392b26STheodore Ts'o prefetch_ios); 33373d392b26STheodore Ts'o trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, 33383d392b26STheodore Ts'o prefetch_ios); 33393d392b26STheodore Ts'o if (group >= elr->lr_next_group) { 33403d392b26STheodore Ts'o ret = 1; 33413d392b26STheodore Ts'o if (elr->lr_first_not_zeroed != ngroups && 33423d392b26STheodore Ts'o !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) { 33433d392b26STheodore Ts'o elr->lr_next_group = elr->lr_first_not_zeroed; 33443d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_ITABLE; 33453d392b26STheodore Ts'o ret = 0; 33463d392b26STheodore Ts'o } 33473d392b26STheodore Ts'o } 33483d392b26STheodore Ts'o return ret; 33493d392b26STheodore Ts'o } 3350bfff6873SLukas Czerner 33513d392b26STheodore Ts'o for (; group < ngroups; group++) { 3352bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3353bfff6873SLukas Czerner if (!gdp) { 3354bfff6873SLukas Czerner ret = 1; 3355bfff6873SLukas Czerner break; 3356bfff6873SLukas Czerner } 3357bfff6873SLukas Czerner 3358bfff6873SLukas Czerner if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3359bfff6873SLukas Czerner break; 3360bfff6873SLukas Czerner } 3361bfff6873SLukas Czerner 33627f511862STheodore Ts'o if (group >= ngroups) 3363bfff6873SLukas Czerner ret = 1; 3364bfff6873SLukas Czerner 3365bfff6873SLukas Czerner if (!ret) { 3366bfff6873SLukas Czerner timeout = jiffies; 3367bfff6873SLukas Czerner ret = ext4_init_inode_table(sb, group, 3368bfff6873SLukas Czerner elr->lr_timeout ? 0 : 1); 33693d392b26STheodore Ts'o trace_ext4_lazy_itable_init(sb, group); 3370bfff6873SLukas Czerner if (elr->lr_timeout == 0) { 337151ce6511SLukas Czerner timeout = (jiffies - timeout) * 33723d392b26STheodore Ts'o EXT4_SB(elr->lr_super)->s_li_wait_mult; 3373bfff6873SLukas Czerner elr->lr_timeout = timeout; 3374bfff6873SLukas Czerner } 3375bfff6873SLukas Czerner elr->lr_next_sched = jiffies + elr->lr_timeout; 3376bfff6873SLukas Czerner elr->lr_next_group = group + 1; 3377bfff6873SLukas Czerner } 3378bfff6873SLukas Czerner return ret; 3379bfff6873SLukas Czerner } 3380bfff6873SLukas Czerner 3381bfff6873SLukas Czerner /* 3382bfff6873SLukas Czerner * Remove lr_request from the list_request and free the 33834ed5c033SLukas Czerner * request structure. Should be called with li_list_mtx held 3384bfff6873SLukas Czerner */ 3385bfff6873SLukas Czerner static void ext4_remove_li_request(struct ext4_li_request *elr) 3386bfff6873SLukas Czerner { 3387bfff6873SLukas Czerner if (!elr) 3388bfff6873SLukas Czerner return; 3389bfff6873SLukas Czerner 3390bfff6873SLukas Czerner list_del(&elr->lr_request); 33913d392b26STheodore Ts'o EXT4_SB(elr->lr_super)->s_li_request = NULL; 3392bfff6873SLukas Czerner kfree(elr); 3393bfff6873SLukas Czerner } 3394bfff6873SLukas Czerner 3395bfff6873SLukas Czerner static void ext4_unregister_li_request(struct super_block *sb) 3396bfff6873SLukas Czerner { 33971bb933fbSLukas Czerner mutex_lock(&ext4_li_mtx); 33981bb933fbSLukas Czerner if (!ext4_li_info) { 33991bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3400bfff6873SLukas Czerner return; 34011bb933fbSLukas Czerner } 3402bfff6873SLukas Czerner 3403bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 34041bb933fbSLukas Czerner ext4_remove_li_request(EXT4_SB(sb)->s_li_request); 3405bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 34061bb933fbSLukas Czerner mutex_unlock(&ext4_li_mtx); 3407bfff6873SLukas Czerner } 3408bfff6873SLukas Czerner 34098f1f7453SEric Sandeen static struct task_struct *ext4_lazyinit_task; 34108f1f7453SEric Sandeen 3411bfff6873SLukas Czerner /* 3412bfff6873SLukas Czerner * This is the function where ext4lazyinit thread lives. It walks 3413bfff6873SLukas Czerner * through the request list searching for next scheduled filesystem. 3414bfff6873SLukas Czerner * When such a fs is found, run the lazy initialization request 3415bfff6873SLukas Czerner * (ext4_rn_li_request) and keep track of the time spend in this 3416bfff6873SLukas Czerner * function. Based on that time we compute next schedule time of 3417bfff6873SLukas Czerner * the request. When walking through the list is complete, compute 3418bfff6873SLukas Czerner * next waking time and put itself into sleep. 3419bfff6873SLukas Czerner */ 3420bfff6873SLukas Czerner static int ext4_lazyinit_thread(void *arg) 3421bfff6873SLukas Czerner { 3422bfff6873SLukas Czerner struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg; 3423bfff6873SLukas Czerner struct list_head *pos, *n; 3424bfff6873SLukas Czerner struct ext4_li_request *elr; 34254ed5c033SLukas Czerner unsigned long next_wakeup, cur; 3426bfff6873SLukas Czerner 3427bfff6873SLukas Czerner BUG_ON(NULL == eli); 3428bfff6873SLukas Czerner 3429bfff6873SLukas Czerner cont_thread: 3430bfff6873SLukas Czerner while (true) { 3431bfff6873SLukas Czerner next_wakeup = MAX_JIFFY_OFFSET; 3432bfff6873SLukas Czerner 3433bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3434bfff6873SLukas Czerner if (list_empty(&eli->li_request_list)) { 3435bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3436bfff6873SLukas Czerner goto exit_thread; 3437bfff6873SLukas Czerner } 3438bfff6873SLukas Czerner list_for_each_safe(pos, n, &eli->li_request_list) { 3439e22834f0SDmitry Monakhov int err = 0; 3440e22834f0SDmitry Monakhov int progress = 0; 3441bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3442bfff6873SLukas Czerner lr_request); 3443bfff6873SLukas Czerner 3444e22834f0SDmitry Monakhov if (time_before(jiffies, elr->lr_next_sched)) { 3445e22834f0SDmitry Monakhov if (time_before(elr->lr_next_sched, next_wakeup)) 3446e22834f0SDmitry Monakhov next_wakeup = elr->lr_next_sched; 3447e22834f0SDmitry Monakhov continue; 3448e22834f0SDmitry Monakhov } 3449e22834f0SDmitry Monakhov if (down_read_trylock(&elr->lr_super->s_umount)) { 3450e22834f0SDmitry Monakhov if (sb_start_write_trylock(elr->lr_super)) { 3451e22834f0SDmitry Monakhov progress = 1; 3452e22834f0SDmitry Monakhov /* 3453e22834f0SDmitry Monakhov * We hold sb->s_umount, sb can not 3454e22834f0SDmitry Monakhov * be removed from the list, it is 3455e22834f0SDmitry Monakhov * now safe to drop li_list_mtx 3456e22834f0SDmitry Monakhov */ 3457e22834f0SDmitry Monakhov mutex_unlock(&eli->li_list_mtx); 3458e22834f0SDmitry Monakhov err = ext4_run_li_request(elr); 3459e22834f0SDmitry Monakhov sb_end_write(elr->lr_super); 3460e22834f0SDmitry Monakhov mutex_lock(&eli->li_list_mtx); 3461e22834f0SDmitry Monakhov n = pos->next; 3462e22834f0SDmitry Monakhov } 3463e22834f0SDmitry Monakhov up_read((&elr->lr_super->s_umount)); 3464e22834f0SDmitry Monakhov } 3465b2c78cd0STheodore Ts'o /* error, remove the lazy_init job */ 3466e22834f0SDmitry Monakhov if (err) { 3467bfff6873SLukas Czerner ext4_remove_li_request(elr); 3468bfff6873SLukas Czerner continue; 3469bfff6873SLukas Czerner } 3470e22834f0SDmitry Monakhov if (!progress) { 3471e22834f0SDmitry Monakhov elr->lr_next_sched = jiffies + 3472e22834f0SDmitry Monakhov (prandom_u32() 3473e22834f0SDmitry Monakhov % (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3474b2c78cd0STheodore Ts'o } 3475bfff6873SLukas Czerner if (time_before(elr->lr_next_sched, next_wakeup)) 3476bfff6873SLukas Czerner next_wakeup = elr->lr_next_sched; 3477bfff6873SLukas Czerner } 3478bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3479bfff6873SLukas Czerner 3480a0acae0eSTejun Heo try_to_freeze(); 3481bfff6873SLukas Czerner 34824ed5c033SLukas Czerner cur = jiffies; 34834ed5c033SLukas Czerner if ((time_after_eq(cur, next_wakeup)) || 3484f4245bd4SLukas Czerner (MAX_JIFFY_OFFSET == next_wakeup)) { 3485bfff6873SLukas Czerner cond_resched(); 3486bfff6873SLukas Czerner continue; 3487bfff6873SLukas Czerner } 3488bfff6873SLukas Czerner 34894ed5c033SLukas Czerner schedule_timeout_interruptible(next_wakeup - cur); 34904ed5c033SLukas Czerner 34918f1f7453SEric Sandeen if (kthread_should_stop()) { 34928f1f7453SEric Sandeen ext4_clear_request_list(); 34938f1f7453SEric Sandeen goto exit_thread; 34948f1f7453SEric Sandeen } 3495bfff6873SLukas Czerner } 3496bfff6873SLukas Czerner 3497bfff6873SLukas Czerner exit_thread: 3498bfff6873SLukas Czerner /* 3499bfff6873SLukas Czerner * It looks like the request list is empty, but we need 3500bfff6873SLukas Czerner * to check it under the li_list_mtx lock, to prevent any 3501bfff6873SLukas Czerner * additions into it, and of course we should lock ext4_li_mtx 3502bfff6873SLukas Czerner * to atomically free the list and ext4_li_info, because at 3503bfff6873SLukas Czerner * this point another ext4 filesystem could be registering 3504bfff6873SLukas Czerner * new one. 3505bfff6873SLukas Czerner */ 3506bfff6873SLukas Czerner mutex_lock(&ext4_li_mtx); 3507bfff6873SLukas Czerner mutex_lock(&eli->li_list_mtx); 3508bfff6873SLukas Czerner if (!list_empty(&eli->li_request_list)) { 3509bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3510bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3511bfff6873SLukas Czerner goto cont_thread; 3512bfff6873SLukas Czerner } 3513bfff6873SLukas Czerner mutex_unlock(&eli->li_list_mtx); 3514bfff6873SLukas Czerner kfree(ext4_li_info); 3515bfff6873SLukas Czerner ext4_li_info = NULL; 3516bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3517bfff6873SLukas Czerner 3518bfff6873SLukas Czerner return 0; 3519bfff6873SLukas Czerner } 3520bfff6873SLukas Czerner 3521bfff6873SLukas Czerner static void ext4_clear_request_list(void) 3522bfff6873SLukas Czerner { 3523bfff6873SLukas Czerner struct list_head *pos, *n; 3524bfff6873SLukas Czerner struct ext4_li_request *elr; 3525bfff6873SLukas Czerner 3526bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3527bfff6873SLukas Czerner list_for_each_safe(pos, n, &ext4_li_info->li_request_list) { 3528bfff6873SLukas Czerner elr = list_entry(pos, struct ext4_li_request, 3529bfff6873SLukas Czerner lr_request); 3530bfff6873SLukas Czerner ext4_remove_li_request(elr); 3531bfff6873SLukas Czerner } 3532bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3533bfff6873SLukas Czerner } 3534bfff6873SLukas Czerner 3535bfff6873SLukas Czerner static int ext4_run_lazyinit_thread(void) 3536bfff6873SLukas Czerner { 35378f1f7453SEric Sandeen ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread, 35388f1f7453SEric Sandeen ext4_li_info, "ext4lazyinit"); 35398f1f7453SEric Sandeen if (IS_ERR(ext4_lazyinit_task)) { 35408f1f7453SEric Sandeen int err = PTR_ERR(ext4_lazyinit_task); 3541bfff6873SLukas Czerner ext4_clear_request_list(); 3542bfff6873SLukas Czerner kfree(ext4_li_info); 3543bfff6873SLukas Czerner ext4_li_info = NULL; 354492b97816STheodore Ts'o printk(KERN_CRIT "EXT4-fs: error %d creating inode table " 3545bfff6873SLukas Czerner "initialization thread\n", 3546bfff6873SLukas Czerner err); 3547bfff6873SLukas Czerner return err; 3548bfff6873SLukas Czerner } 3549bfff6873SLukas Czerner ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING; 3550bfff6873SLukas Czerner return 0; 3551bfff6873SLukas Czerner } 3552bfff6873SLukas Czerner 3553bfff6873SLukas Czerner /* 3554bfff6873SLukas Czerner * Check whether it make sense to run itable init. thread or not. 3555bfff6873SLukas Czerner * If there is at least one uninitialized inode table, return 3556bfff6873SLukas Czerner * corresponding group number, else the loop goes through all 3557bfff6873SLukas Czerner * groups and return total number of groups. 3558bfff6873SLukas Czerner */ 3559bfff6873SLukas Czerner static ext4_group_t ext4_has_uninit_itable(struct super_block *sb) 3560bfff6873SLukas Czerner { 3561bfff6873SLukas Czerner ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count; 3562bfff6873SLukas Czerner struct ext4_group_desc *gdp = NULL; 3563bfff6873SLukas Czerner 35648844618dSTheodore Ts'o if (!ext4_has_group_desc_csum(sb)) 35658844618dSTheodore Ts'o return ngroups; 35668844618dSTheodore Ts'o 3567bfff6873SLukas Czerner for (group = 0; group < ngroups; group++) { 3568bfff6873SLukas Czerner gdp = ext4_get_group_desc(sb, group, NULL); 3569bfff6873SLukas Czerner if (!gdp) 3570bfff6873SLukas Czerner continue; 3571bfff6873SLukas Czerner 357250122847STheodore Ts'o if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))) 3573bfff6873SLukas Czerner break; 3574bfff6873SLukas Czerner } 3575bfff6873SLukas Czerner 3576bfff6873SLukas Czerner return group; 3577bfff6873SLukas Czerner } 3578bfff6873SLukas Czerner 3579bfff6873SLukas Czerner static int ext4_li_info_new(void) 3580bfff6873SLukas Czerner { 3581bfff6873SLukas Czerner struct ext4_lazy_init *eli = NULL; 3582bfff6873SLukas Czerner 3583bfff6873SLukas Czerner eli = kzalloc(sizeof(*eli), GFP_KERNEL); 3584bfff6873SLukas Czerner if (!eli) 3585bfff6873SLukas Czerner return -ENOMEM; 3586bfff6873SLukas Czerner 3587bfff6873SLukas Czerner INIT_LIST_HEAD(&eli->li_request_list); 3588bfff6873SLukas Czerner mutex_init(&eli->li_list_mtx); 3589bfff6873SLukas Czerner 3590bfff6873SLukas Czerner eli->li_state |= EXT4_LAZYINIT_QUIT; 3591bfff6873SLukas Czerner 3592bfff6873SLukas Czerner ext4_li_info = eli; 3593bfff6873SLukas Czerner 3594bfff6873SLukas Czerner return 0; 3595bfff6873SLukas Czerner } 3596bfff6873SLukas Czerner 3597bfff6873SLukas Czerner static struct ext4_li_request *ext4_li_request_new(struct super_block *sb, 3598bfff6873SLukas Czerner ext4_group_t start) 3599bfff6873SLukas Czerner { 3600bfff6873SLukas Czerner struct ext4_li_request *elr; 3601bfff6873SLukas Czerner 3602bfff6873SLukas Czerner elr = kzalloc(sizeof(*elr), GFP_KERNEL); 3603bfff6873SLukas Czerner if (!elr) 3604bfff6873SLukas Czerner return NULL; 3605bfff6873SLukas Czerner 3606bfff6873SLukas Czerner elr->lr_super = sb; 36073d392b26STheodore Ts'o elr->lr_first_not_zeroed = start; 36083d392b26STheodore Ts'o if (test_opt(sb, PREFETCH_BLOCK_BITMAPS)) 36093d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP; 36103d392b26STheodore Ts'o else { 36113d392b26STheodore Ts'o elr->lr_mode = EXT4_LI_MODE_ITABLE; 3612bfff6873SLukas Czerner elr->lr_next_group = start; 36133d392b26STheodore Ts'o } 3614bfff6873SLukas Czerner 3615bfff6873SLukas Czerner /* 3616bfff6873SLukas Czerner * Randomize first schedule time of the request to 3617bfff6873SLukas Czerner * spread the inode table initialization requests 3618bfff6873SLukas Czerner * better. 3619bfff6873SLukas Czerner */ 3620dd1f723bSTheodore Ts'o elr->lr_next_sched = jiffies + (prandom_u32() % 3621dd1f723bSTheodore Ts'o (EXT4_DEF_LI_MAX_START_DELAY * HZ)); 3622bfff6873SLukas Czerner return elr; 3623bfff6873SLukas Czerner } 3624bfff6873SLukas Czerner 36257f511862STheodore Ts'o int ext4_register_li_request(struct super_block *sb, 3626bfff6873SLukas Czerner ext4_group_t first_not_zeroed) 3627bfff6873SLukas Czerner { 3628bfff6873SLukas Czerner struct ext4_sb_info *sbi = EXT4_SB(sb); 36297f511862STheodore Ts'o struct ext4_li_request *elr = NULL; 363049598e04SJun Piao ext4_group_t ngroups = sbi->s_groups_count; 36316c5a6cb9SAndrew Morton int ret = 0; 3632bfff6873SLukas Czerner 36337f511862STheodore Ts'o mutex_lock(&ext4_li_mtx); 363451ce6511SLukas Czerner if (sbi->s_li_request != NULL) { 363551ce6511SLukas Czerner /* 363651ce6511SLukas Czerner * Reset timeout so it can be computed again, because 363751ce6511SLukas Czerner * s_li_wait_mult might have changed. 363851ce6511SLukas Czerner */ 363951ce6511SLukas Czerner sbi->s_li_request->lr_timeout = 0; 36407f511862STheodore Ts'o goto out; 364151ce6511SLukas Czerner } 3642bfff6873SLukas Czerner 36433d392b26STheodore Ts'o if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) && 36443d392b26STheodore Ts'o (first_not_zeroed == ngroups || sb_rdonly(sb) || 36453d392b26STheodore Ts'o !test_opt(sb, INIT_INODE_TABLE))) 36467f511862STheodore Ts'o goto out; 3647bfff6873SLukas Czerner 3648bfff6873SLukas Czerner elr = ext4_li_request_new(sb, first_not_zeroed); 36497f511862STheodore Ts'o if (!elr) { 36507f511862STheodore Ts'o ret = -ENOMEM; 36517f511862STheodore Ts'o goto out; 36527f511862STheodore Ts'o } 3653bfff6873SLukas Czerner 3654bfff6873SLukas Czerner if (NULL == ext4_li_info) { 3655bfff6873SLukas Czerner ret = ext4_li_info_new(); 3656bfff6873SLukas Czerner if (ret) 3657bfff6873SLukas Czerner goto out; 3658bfff6873SLukas Czerner } 3659bfff6873SLukas Czerner 3660bfff6873SLukas Czerner mutex_lock(&ext4_li_info->li_list_mtx); 3661bfff6873SLukas Czerner list_add(&elr->lr_request, &ext4_li_info->li_request_list); 3662bfff6873SLukas Czerner mutex_unlock(&ext4_li_info->li_list_mtx); 3663bfff6873SLukas Czerner 3664bfff6873SLukas Czerner sbi->s_li_request = elr; 366546e4690bSTao Ma /* 366646e4690bSTao Ma * set elr to NULL here since it has been inserted to 366746e4690bSTao Ma * the request_list and the removal and free of it is 366846e4690bSTao Ma * handled by ext4_clear_request_list from now on. 366946e4690bSTao Ma */ 367046e4690bSTao Ma elr = NULL; 3671bfff6873SLukas Czerner 3672bfff6873SLukas Czerner if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) { 3673bfff6873SLukas Czerner ret = ext4_run_lazyinit_thread(); 3674bfff6873SLukas Czerner if (ret) 3675bfff6873SLukas Czerner goto out; 3676bfff6873SLukas Czerner } 3677bfff6873SLukas Czerner out: 3678bfff6873SLukas Czerner mutex_unlock(&ext4_li_mtx); 3679beed5ecbSNicolas Kaiser if (ret) 3680bfff6873SLukas Czerner kfree(elr); 3681bfff6873SLukas Czerner return ret; 3682bfff6873SLukas Czerner } 3683bfff6873SLukas Czerner 3684bfff6873SLukas Czerner /* 3685bfff6873SLukas Czerner * We do not need to lock anything since this is called on 3686bfff6873SLukas Czerner * module unload. 3687bfff6873SLukas Czerner */ 3688bfff6873SLukas Czerner static void ext4_destroy_lazyinit_thread(void) 3689bfff6873SLukas Czerner { 3690bfff6873SLukas Czerner /* 3691bfff6873SLukas Czerner * If thread exited earlier 3692bfff6873SLukas Czerner * there's nothing to be done. 3693bfff6873SLukas Czerner */ 36948f1f7453SEric Sandeen if (!ext4_li_info || !ext4_lazyinit_task) 3695bfff6873SLukas Czerner return; 3696bfff6873SLukas Czerner 36978f1f7453SEric Sandeen kthread_stop(ext4_lazyinit_task); 3698bfff6873SLukas Czerner } 3699bfff6873SLukas Czerner 370025ed6e8aSDarrick J. Wong static int set_journal_csum_feature_set(struct super_block *sb) 370125ed6e8aSDarrick J. Wong { 370225ed6e8aSDarrick J. Wong int ret = 1; 370325ed6e8aSDarrick J. Wong int compat, incompat; 370425ed6e8aSDarrick J. Wong struct ext4_sb_info *sbi = EXT4_SB(sb); 370525ed6e8aSDarrick J. Wong 37069aa5d32bSDmitry Monakhov if (ext4_has_metadata_csum(sb)) { 3707db9ee220SDarrick J. Wong /* journal checksum v3 */ 370825ed6e8aSDarrick J. Wong compat = 0; 3709db9ee220SDarrick J. Wong incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3; 371025ed6e8aSDarrick J. Wong } else { 371125ed6e8aSDarrick J. Wong /* journal checksum v1 */ 371225ed6e8aSDarrick J. Wong compat = JBD2_FEATURE_COMPAT_CHECKSUM; 371325ed6e8aSDarrick J. Wong incompat = 0; 371425ed6e8aSDarrick J. Wong } 371525ed6e8aSDarrick J. Wong 3716feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 3717feb8c6d3SDarrick J. Wong JBD2_FEATURE_COMPAT_CHECKSUM, 0, 3718feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V3 | 3719feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_CSUM_V2); 372025ed6e8aSDarrick J. Wong if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 372125ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 372225ed6e8aSDarrick J. Wong compat, 0, 372325ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT | 372425ed6e8aSDarrick J. Wong incompat); 372525ed6e8aSDarrick J. Wong } else if (test_opt(sb, JOURNAL_CHECKSUM)) { 372625ed6e8aSDarrick J. Wong ret = jbd2_journal_set_features(sbi->s_journal, 372725ed6e8aSDarrick J. Wong compat, 0, 372825ed6e8aSDarrick J. Wong incompat); 372925ed6e8aSDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 373025ed6e8aSDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 373125ed6e8aSDarrick J. Wong } else { 3732feb8c6d3SDarrick J. Wong jbd2_journal_clear_features(sbi->s_journal, 0, 0, 3733feb8c6d3SDarrick J. Wong JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT); 373425ed6e8aSDarrick J. Wong } 373525ed6e8aSDarrick J. Wong 373625ed6e8aSDarrick J. Wong return ret; 373725ed6e8aSDarrick J. Wong } 373825ed6e8aSDarrick J. Wong 3739952fc18eSTheodore Ts'o /* 3740952fc18eSTheodore Ts'o * Note: calculating the overhead so we can be compatible with 3741952fc18eSTheodore Ts'o * historical BSD practice is quite difficult in the face of 3742952fc18eSTheodore Ts'o * clusters/bigalloc. This is because multiple metadata blocks from 3743952fc18eSTheodore Ts'o * different block group can end up in the same allocation cluster. 3744952fc18eSTheodore Ts'o * Calculating the exact overhead in the face of clustered allocation 3745952fc18eSTheodore Ts'o * requires either O(all block bitmaps) in memory or O(number of block 3746952fc18eSTheodore Ts'o * groups**2) in time. We will still calculate the superblock for 3747952fc18eSTheodore Ts'o * older file systems --- and if we come across with a bigalloc file 3748952fc18eSTheodore Ts'o * system with zero in s_overhead_clusters the estimate will be close to 3749952fc18eSTheodore Ts'o * correct especially for very large cluster sizes --- but for newer 3750952fc18eSTheodore Ts'o * file systems, it's better to calculate this figure once at mkfs 3751952fc18eSTheodore Ts'o * time, and store it in the superblock. If the superblock value is 3752952fc18eSTheodore Ts'o * present (even for non-bigalloc file systems), we will use it. 3753952fc18eSTheodore Ts'o */ 3754952fc18eSTheodore Ts'o static int count_overhead(struct super_block *sb, ext4_group_t grp, 3755952fc18eSTheodore Ts'o char *buf) 3756952fc18eSTheodore Ts'o { 3757952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3758952fc18eSTheodore Ts'o struct ext4_group_desc *gdp; 3759952fc18eSTheodore Ts'o ext4_fsblk_t first_block, last_block, b; 3760952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3761952fc18eSTheodore Ts'o int s, j, count = 0; 3762952fc18eSTheodore Ts'o 3763e2b911c5SDarrick J. Wong if (!ext4_has_feature_bigalloc(sb)) 37640548bbb8STheodore Ts'o return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) + 37650548bbb8STheodore Ts'o sbi->s_itb_per_group + 2); 37660548bbb8STheodore Ts'o 3767952fc18eSTheodore Ts'o first_block = le32_to_cpu(sbi->s_es->s_first_data_block) + 3768952fc18eSTheodore Ts'o (grp * EXT4_BLOCKS_PER_GROUP(sb)); 3769952fc18eSTheodore Ts'o last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1; 3770952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3771952fc18eSTheodore Ts'o gdp = ext4_get_group_desc(sb, i, NULL); 3772952fc18eSTheodore Ts'o b = ext4_block_bitmap(sb, gdp); 3773952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3774952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3775952fc18eSTheodore Ts'o count++; 3776952fc18eSTheodore Ts'o } 3777952fc18eSTheodore Ts'o b = ext4_inode_bitmap(sb, gdp); 3778952fc18eSTheodore Ts'o if (b >= first_block && b <= last_block) { 3779952fc18eSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf); 3780952fc18eSTheodore Ts'o count++; 3781952fc18eSTheodore Ts'o } 3782952fc18eSTheodore Ts'o b = ext4_inode_table(sb, gdp); 3783952fc18eSTheodore Ts'o if (b >= first_block && b + sbi->s_itb_per_group <= last_block) 3784952fc18eSTheodore Ts'o for (j = 0; j < sbi->s_itb_per_group; j++, b++) { 3785952fc18eSTheodore Ts'o int c = EXT4_B2C(sbi, b - first_block); 3786952fc18eSTheodore Ts'o ext4_set_bit(c, buf); 3787952fc18eSTheodore Ts'o count++; 3788952fc18eSTheodore Ts'o } 3789952fc18eSTheodore Ts'o if (i != grp) 3790952fc18eSTheodore Ts'o continue; 3791952fc18eSTheodore Ts'o s = 0; 3792952fc18eSTheodore Ts'o if (ext4_bg_has_super(sb, grp)) { 3793952fc18eSTheodore Ts'o ext4_set_bit(s++, buf); 3794952fc18eSTheodore Ts'o count++; 3795952fc18eSTheodore Ts'o } 3796c48ae41bSTheodore Ts'o j = ext4_bg_num_gdb(sb, grp); 3797c48ae41bSTheodore Ts'o if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) { 3798c48ae41bSTheodore Ts'o ext4_error(sb, "Invalid number of block group " 3799c48ae41bSTheodore Ts'o "descriptor blocks: %d", j); 3800c48ae41bSTheodore Ts'o j = EXT4_BLOCKS_PER_GROUP(sb) - s; 3801952fc18eSTheodore Ts'o } 3802c48ae41bSTheodore Ts'o count += j; 3803c48ae41bSTheodore Ts'o for (; j > 0; j--) 3804c48ae41bSTheodore Ts'o ext4_set_bit(EXT4_B2C(sbi, s++), buf); 3805952fc18eSTheodore Ts'o } 3806952fc18eSTheodore Ts'o if (!count) 3807952fc18eSTheodore Ts'o return 0; 3808952fc18eSTheodore Ts'o return EXT4_CLUSTERS_PER_GROUP(sb) - 3809952fc18eSTheodore Ts'o ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8); 3810952fc18eSTheodore Ts'o } 3811952fc18eSTheodore Ts'o 3812952fc18eSTheodore Ts'o /* 3813952fc18eSTheodore Ts'o * Compute the overhead and stash it in sbi->s_overhead 3814952fc18eSTheodore Ts'o */ 3815952fc18eSTheodore Ts'o int ext4_calculate_overhead(struct super_block *sb) 3816952fc18eSTheodore Ts'o { 3817952fc18eSTheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 3818952fc18eSTheodore Ts'o struct ext4_super_block *es = sbi->s_es; 38193c816dedSEric Whitney struct inode *j_inode; 38203c816dedSEric Whitney unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum); 3821952fc18eSTheodore Ts'o ext4_group_t i, ngroups = ext4_get_groups_count(sb); 3822952fc18eSTheodore Ts'o ext4_fsblk_t overhead = 0; 38234fdb5543SDmitry Monakhov char *buf = (char *) get_zeroed_page(GFP_NOFS); 3824952fc18eSTheodore Ts'o 3825952fc18eSTheodore Ts'o if (!buf) 3826952fc18eSTheodore Ts'o return -ENOMEM; 3827952fc18eSTheodore Ts'o 3828952fc18eSTheodore Ts'o /* 3829952fc18eSTheodore Ts'o * Compute the overhead (FS structures). This is constant 3830952fc18eSTheodore Ts'o * for a given filesystem unless the number of block groups 3831952fc18eSTheodore Ts'o * changes so we cache the previous value until it does. 3832952fc18eSTheodore Ts'o */ 3833952fc18eSTheodore Ts'o 3834952fc18eSTheodore Ts'o /* 3835952fc18eSTheodore Ts'o * All of the blocks before first_data_block are overhead 3836952fc18eSTheodore Ts'o */ 3837952fc18eSTheodore Ts'o overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block)); 3838952fc18eSTheodore Ts'o 3839952fc18eSTheodore Ts'o /* 3840952fc18eSTheodore Ts'o * Add the overhead found in each block group 3841952fc18eSTheodore Ts'o */ 3842952fc18eSTheodore Ts'o for (i = 0; i < ngroups; i++) { 3843952fc18eSTheodore Ts'o int blks; 3844952fc18eSTheodore Ts'o 3845952fc18eSTheodore Ts'o blks = count_overhead(sb, i, buf); 3846952fc18eSTheodore Ts'o overhead += blks; 3847952fc18eSTheodore Ts'o if (blks) 3848952fc18eSTheodore Ts'o memset(buf, 0, PAGE_SIZE); 3849952fc18eSTheodore Ts'o cond_resched(); 3850952fc18eSTheodore Ts'o } 38513c816dedSEric Whitney 38523c816dedSEric Whitney /* 38533c816dedSEric Whitney * Add the internal journal blocks whether the journal has been 38543c816dedSEric Whitney * loaded or not 38553c816dedSEric Whitney */ 3856ee7ed3aaSChunguang Xu if (sbi->s_journal && !sbi->s_journal_bdev) 3857810da240SLukas Czerner overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen); 3858f1eec3b0SRitesh Harjani else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) { 3859f1eec3b0SRitesh Harjani /* j_inum for internal journal is non-zero */ 38603c816dedSEric Whitney j_inode = ext4_get_journal_inode(sb, j_inum); 38613c816dedSEric Whitney if (j_inode) { 38623c816dedSEric Whitney j_blocks = j_inode->i_size >> sb->s_blocksize_bits; 38633c816dedSEric Whitney overhead += EXT4_NUM_B2C(sbi, j_blocks); 38643c816dedSEric Whitney iput(j_inode); 38653c816dedSEric Whitney } else { 38663c816dedSEric Whitney ext4_msg(sb, KERN_ERR, "can't get journal size"); 38673c816dedSEric Whitney } 38683c816dedSEric Whitney } 3869952fc18eSTheodore Ts'o sbi->s_overhead = overhead; 3870952fc18eSTheodore Ts'o smp_wmb(); 3871952fc18eSTheodore Ts'o free_page((unsigned long) buf); 3872952fc18eSTheodore Ts'o return 0; 3873952fc18eSTheodore Ts'o } 3874952fc18eSTheodore Ts'o 3875b5799018STheodore Ts'o static void ext4_set_resv_clusters(struct super_block *sb) 387627dd4385SLukas Czerner { 387727dd4385SLukas Czerner ext4_fsblk_t resv_clusters; 3878b5799018STheodore Ts'o struct ext4_sb_info *sbi = EXT4_SB(sb); 387927dd4385SLukas Czerner 388027dd4385SLukas Czerner /* 388130fac0f7SJan Kara * There's no need to reserve anything when we aren't using extents. 388230fac0f7SJan Kara * The space estimates are exact, there are no unwritten extents, 388330fac0f7SJan Kara * hole punching doesn't need new metadata... This is needed especially 388430fac0f7SJan Kara * to keep ext2/3 backward compatibility. 388530fac0f7SJan Kara */ 3886e2b911c5SDarrick J. Wong if (!ext4_has_feature_extents(sb)) 3887b5799018STheodore Ts'o return; 388830fac0f7SJan Kara /* 388927dd4385SLukas Czerner * By default we reserve 2% or 4096 clusters, whichever is smaller. 389027dd4385SLukas Czerner * This should cover the situations where we can not afford to run 389127dd4385SLukas Czerner * out of space like for example punch hole, or converting 3892556615dcSLukas Czerner * unwritten extents in delalloc path. In most cases such 389327dd4385SLukas Czerner * allocation would require 1, or 2 blocks, higher numbers are 389427dd4385SLukas Czerner * very rare. 389527dd4385SLukas Czerner */ 3896b5799018STheodore Ts'o resv_clusters = (ext4_blocks_count(sbi->s_es) >> 3897b5799018STheodore Ts'o sbi->s_cluster_bits); 389827dd4385SLukas Czerner 389927dd4385SLukas Czerner do_div(resv_clusters, 50); 390027dd4385SLukas Czerner resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096); 390127dd4385SLukas Czerner 3902b5799018STheodore Ts'o atomic64_set(&sbi->s_resv_clusters, resv_clusters); 390327dd4385SLukas Czerner } 390427dd4385SLukas Czerner 3905617ba13bSMingming Cao static int ext4_fill_super(struct super_block *sb, void *data, int silent) 3906ac27a0ecSDave Kleikamp { 39075e405595SDan Williams struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev); 3908d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 39091d0c3924STheodore Ts'o struct buffer_head *bh, **group_desc; 3910617ba13bSMingming Cao struct ext4_super_block *es = NULL; 39115aee0f8aSTheodore Ts'o struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 39127c990728SSuraj Jitindar Singh struct flex_groups **flex_groups; 3913617ba13bSMingming Cao ext4_fsblk_t block; 3914617ba13bSMingming Cao ext4_fsblk_t sb_block = get_sb_block(&data); 391570bbb3e0SAndrew Morton ext4_fsblk_t logical_sb_block; 3916ac27a0ecSDave Kleikamp unsigned long offset = 0; 3917ac27a0ecSDave Kleikamp unsigned long journal_devnum = 0; 3918ac27a0ecSDave Kleikamp unsigned long def_mount_opts; 3919ac27a0ecSDave Kleikamp struct inode *root; 39200390131bSFrank Mayhar const char *descr; 3921dcc7dae3SCyrill Gorcunov int ret = -ENOMEM; 3922281b5995STheodore Ts'o int blocksize, clustersize; 39234ec11028STheodore Ts'o unsigned int db_count; 39244ec11028STheodore Ts'o unsigned int i; 3925ef5fd681SKaixu Xia int needs_recovery, has_huge_files; 3926bd81d8eeSLaurent Vivier __u64 blocks_count; 392707aa2ea1SLukas Czerner int err = 0; 3928b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 3929bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 3930ac27a0ecSDave Kleikamp 39315aee0f8aSTheodore Ts'o if ((data && !orig_data) || !sbi) 39325aee0f8aSTheodore Ts'o goto out_free_base; 3933705895b6SPekka Enberg 3934aed9eb1bSColin Ian King sbi->s_daxdev = dax_dev; 3935705895b6SPekka Enberg sbi->s_blockgroup_lock = 3936705895b6SPekka Enberg kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 39375aee0f8aSTheodore Ts'o if (!sbi->s_blockgroup_lock) 39385aee0f8aSTheodore Ts'o goto out_free_base; 39395aee0f8aSTheodore Ts'o 3940ac27a0ecSDave Kleikamp sb->s_fs_info = sbi; 39412c0544b2STheodore Ts'o sbi->s_sb = sb; 3942240799cdSTheodore Ts'o sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3943d9c9bef1SMiklos Szeredi sbi->s_sb_block = sb_block; 3944f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 3945f613dfcbSTheodore Ts'o sbi->s_sectors_written_start = 3946dbae2c55SMichael Callahan part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]); 3947ac27a0ecSDave Kleikamp 39489f6200bbSTheodore Ts'o /* Cleanup superblock name */ 3949ec3904dcSRasmus Villemoes strreplace(sb->s_id, '/', '!'); 39509f6200bbSTheodore Ts'o 395107aa2ea1SLukas Czerner /* -EINVAL is default */ 3952dcc7dae3SCyrill Gorcunov ret = -EINVAL; 3953617ba13bSMingming Cao blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3954ac27a0ecSDave Kleikamp if (!blocksize) { 3955b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3956ac27a0ecSDave Kleikamp goto out_fail; 3957ac27a0ecSDave Kleikamp } 3958ac27a0ecSDave Kleikamp 3959ac27a0ecSDave Kleikamp /* 3960617ba13bSMingming Cao * The ext4 superblock will not be buffer aligned for other than 1kB 3961ac27a0ecSDave Kleikamp * block sizes. We need to calculate the offset from buffer start. 3962ac27a0ecSDave Kleikamp */ 3963617ba13bSMingming Cao if (blocksize != EXT4_MIN_BLOCK_SIZE) { 396470bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 396570bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 3966ac27a0ecSDave Kleikamp } else { 396770bbb3e0SAndrew Morton logical_sb_block = sb_block; 3968ac27a0ecSDave Kleikamp } 3969ac27a0ecSDave Kleikamp 39708394a6abSzhangyi (F) bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 39718394a6abSzhangyi (F) if (IS_ERR(bh)) { 3972b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "unable to read superblock"); 39738394a6abSzhangyi (F) ret = PTR_ERR(bh); 39748394a6abSzhangyi (F) bh = NULL; 3975ac27a0ecSDave Kleikamp goto out_fail; 3976ac27a0ecSDave Kleikamp } 3977ac27a0ecSDave Kleikamp /* 3978ac27a0ecSDave Kleikamp * Note: s_es must be initialized as soon as possible because 3979617ba13bSMingming Cao * some ext4 macro-instructions depend on its value 3980ac27a0ecSDave Kleikamp */ 39812716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 3982ac27a0ecSDave Kleikamp sbi->s_es = es; 3983ac27a0ecSDave Kleikamp sb->s_magic = le16_to_cpu(es->s_magic); 3984617ba13bSMingming Cao if (sb->s_magic != EXT4_SUPER_MAGIC) 3985617ba13bSMingming Cao goto cantfind_ext4; 3986afc32f7eSTheodore Ts'o sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written); 3987ac27a0ecSDave Kleikamp 3988feb0ab32SDarrick J. Wong /* Warn if metadata_csum and gdt_csum are both set. */ 3989e2b911c5SDarrick J. Wong if (ext4_has_feature_metadata_csum(sb) && 3990e2b911c5SDarrick J. Wong ext4_has_feature_gdt_csum(sb)) 3991363307e6SJakub Wilk ext4_warning(sb, "metadata_csum and uninit_bg are " 3992feb0ab32SDarrick J. Wong "redundant flags; please run fsck."); 3993feb0ab32SDarrick J. Wong 3994d25425f8SDarrick J. Wong /* Check for a known checksum algorithm */ 3995d25425f8SDarrick J. Wong if (!ext4_verify_csum_type(sb, es)) { 3996d25425f8SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 3997d25425f8SDarrick J. Wong "unknown checksum algorithm."); 3998d25425f8SDarrick J. Wong silent = 1; 3999d25425f8SDarrick J. Wong goto cantfind_ext4; 4000d25425f8SDarrick J. Wong } 4001d25425f8SDarrick J. Wong 40020441984aSDarrick J. Wong /* Load the checksum driver */ 40030441984aSDarrick J. Wong sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0); 40040441984aSDarrick J. Wong if (IS_ERR(sbi->s_chksum_driver)) { 40050441984aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver."); 40060441984aSDarrick J. Wong ret = PTR_ERR(sbi->s_chksum_driver); 40070441984aSDarrick J. Wong sbi->s_chksum_driver = NULL; 40080441984aSDarrick J. Wong goto failed_mount; 40090441984aSDarrick J. Wong } 40100441984aSDarrick J. Wong 4011a9c47317SDarrick J. Wong /* Check superblock checksum */ 4012a9c47317SDarrick J. Wong if (!ext4_superblock_csum_verify(sb, es)) { 4013a9c47317SDarrick J. Wong ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with " 4014a9c47317SDarrick J. Wong "invalid superblock checksum. Run e2fsck?"); 4015a9c47317SDarrick J. Wong silent = 1; 40166a797d27SDarrick J. Wong ret = -EFSBADCRC; 4017a9c47317SDarrick J. Wong goto cantfind_ext4; 4018a9c47317SDarrick J. Wong } 4019a9c47317SDarrick J. Wong 4020a9c47317SDarrick J. Wong /* Precompute checksum seed for all metadata */ 4021e2b911c5SDarrick J. Wong if (ext4_has_feature_csum_seed(sb)) 40228c81bd8fSDarrick J. Wong sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed); 4023dec214d0STahsin Erdogan else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb)) 4024a9c47317SDarrick J. Wong sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid, 4025a9c47317SDarrick J. Wong sizeof(es->s_uuid)); 4026a9c47317SDarrick J. Wong 4027ac27a0ecSDave Kleikamp /* Set defaults before we parse the mount options */ 4028ac27a0ecSDave Kleikamp def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 4029fd8c37ecSTheodore Ts'o set_opt(sb, INIT_INODE_TABLE); 4030617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_DEBUG) 4031fd8c37ecSTheodore Ts'o set_opt(sb, DEBUG); 403287f26807STheodore Ts'o if (def_mount_opts & EXT4_DEFM_BSDGROUPS) 4033fd8c37ecSTheodore Ts'o set_opt(sb, GRPID); 4034617ba13bSMingming Cao if (def_mount_opts & EXT4_DEFM_UID16) 4035fd8c37ecSTheodore Ts'o set_opt(sb, NO_UID32); 4036ea663336SEric Sandeen /* xattr user namespace & acls are now defaulted on */ 4037fd8c37ecSTheodore Ts'o set_opt(sb, XATTR_USER); 403803010a33STheodore Ts'o #ifdef CONFIG_EXT4_FS_POSIX_ACL 4039fd8c37ecSTheodore Ts'o set_opt(sb, POSIX_ACL); 40402e7842b8SHugh Dickins #endif 404198c1a759SDarrick J. Wong /* don't forget to enable journal_csum when metadata_csum is enabled. */ 404298c1a759SDarrick J. Wong if (ext4_has_metadata_csum(sb)) 404398c1a759SDarrick J. Wong set_opt(sb, JOURNAL_CHECKSUM); 404498c1a759SDarrick J. Wong 4045617ba13bSMingming Cao if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA) 4046fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 4047617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED) 4048fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 4049617ba13bSMingming Cao else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK) 4050fd8c37ecSTheodore Ts'o set_opt(sb, WRITEBACK_DATA); 4051ac27a0ecSDave Kleikamp 4052617ba13bSMingming Cao if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) 4053fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_PANIC); 4054bb4f397aSAneesh Kumar K.V else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) 4055fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_CONT); 4056bb4f397aSAneesh Kumar K.V else 4057fd8c37ecSTheodore Ts'o set_opt(sb, ERRORS_RO); 405845f1a9c3SDarrick J. Wong /* block_validity enabled by default; disable with noblock_validity */ 4059fd8c37ecSTheodore Ts'o set_opt(sb, BLOCK_VALIDITY); 40608b67f04aSTheodore Ts'o if (def_mount_opts & EXT4_DEFM_DISCARD) 4061fd8c37ecSTheodore Ts'o set_opt(sb, DISCARD); 4062ac27a0ecSDave Kleikamp 406308cefc7aSEric W. Biederman sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid)); 406408cefc7aSEric W. Biederman sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid)); 406530773840STheodore Ts'o sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ; 406630773840STheodore Ts'o sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 406730773840STheodore Ts'o sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 4068ac27a0ecSDave Kleikamp 40698b67f04aSTheodore Ts'o if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0) 4070fd8c37ecSTheodore Ts'o set_opt(sb, BARRIER); 4071ac27a0ecSDave Kleikamp 40721e2462f9SMingming Cao /* 4073dd919b98SAneesh Kumar K.V * enable delayed allocation by default 4074dd919b98SAneesh Kumar K.V * Use -o nodelalloc to turn it off 4075dd919b98SAneesh Kumar K.V */ 4076bc0b75f7STheodore Ts'o if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) && 40778b67f04aSTheodore Ts'o ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0)) 4078fd8c37ecSTheodore Ts'o set_opt(sb, DELALLOC); 4079dd919b98SAneesh Kumar K.V 408051ce6511SLukas Czerner /* 408151ce6511SLukas Czerner * set default s_li_wait_mult for lazyinit, for the case there is 408251ce6511SLukas Czerner * no mount option specified. 408351ce6511SLukas Czerner */ 408451ce6511SLukas Czerner sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT; 408551ce6511SLukas Czerner 40864f97a681STheodore Ts'o blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size); 4087626b035bSRitesh Harjani 4088626b035bSRitesh Harjani if (blocksize == PAGE_SIZE) 4089626b035bSRitesh Harjani set_opt(sb, DIOREAD_NOLOCK); 4090626b035bSRitesh Harjani 40914f97a681STheodore Ts'o if (blocksize < EXT4_MIN_BLOCK_SIZE || 40924f97a681STheodore Ts'o blocksize > EXT4_MAX_BLOCK_SIZE) { 40934f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, 40944f97a681STheodore Ts'o "Unsupported filesystem blocksize %d (%d log_block_size)", 40954f97a681STheodore Ts'o blocksize, le32_to_cpu(es->s_log_block_size)); 40964f97a681STheodore Ts'o goto failed_mount; 40974f97a681STheodore Ts'o } 40984f97a681STheodore Ts'o 40999803387cSTheodore Ts'o if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) { 41009803387cSTheodore Ts'o sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE; 41019803387cSTheodore Ts'o sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO; 41029803387cSTheodore Ts'o } else { 41039803387cSTheodore Ts'o sbi->s_inode_size = le16_to_cpu(es->s_inode_size); 41049803387cSTheodore Ts'o sbi->s_first_ino = le32_to_cpu(es->s_first_ino); 41059803387cSTheodore Ts'o if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) { 41069803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid first ino: %u", 41079803387cSTheodore Ts'o sbi->s_first_ino); 41089803387cSTheodore Ts'o goto failed_mount; 41099803387cSTheodore Ts'o } 41109803387cSTheodore Ts'o if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) || 41119803387cSTheodore Ts'o (!is_power_of_2(sbi->s_inode_size)) || 41129803387cSTheodore Ts'o (sbi->s_inode_size > blocksize)) { 41139803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 41149803387cSTheodore Ts'o "unsupported inode size: %d", 41159803387cSTheodore Ts'o sbi->s_inode_size); 41164f97a681STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize); 41179803387cSTheodore Ts'o goto failed_mount; 41189803387cSTheodore Ts'o } 41199803387cSTheodore Ts'o /* 41209803387cSTheodore Ts'o * i_atime_extra is the last extra field available for 41219803387cSTheodore Ts'o * [acm]times in struct ext4_inode. Checking for that 41229803387cSTheodore Ts'o * field should suffice to ensure we have extra space 41239803387cSTheodore Ts'o * for all three. 41249803387cSTheodore Ts'o */ 41259803387cSTheodore Ts'o if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) + 41269803387cSTheodore Ts'o sizeof(((struct ext4_inode *)0)->i_atime_extra)) { 41279803387cSTheodore Ts'o sb->s_time_gran = 1; 41289803387cSTheodore Ts'o sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX; 41299803387cSTheodore Ts'o } else { 41309803387cSTheodore Ts'o sb->s_time_gran = NSEC_PER_SEC; 41319803387cSTheodore Ts'o sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX; 41329803387cSTheodore Ts'o } 41339803387cSTheodore Ts'o sb->s_time_min = EXT4_TIMESTAMP_MIN; 41349803387cSTheodore Ts'o } 41359803387cSTheodore Ts'o if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) { 41369803387cSTheodore Ts'o sbi->s_want_extra_isize = sizeof(struct ext4_inode) - 41379803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE; 41389803387cSTheodore Ts'o if (ext4_has_feature_extra_isize(sb)) { 41399803387cSTheodore Ts'o unsigned v, max = (sbi->s_inode_size - 41409803387cSTheodore Ts'o EXT4_GOOD_OLD_INODE_SIZE); 41419803387cSTheodore Ts'o 41429803387cSTheodore Ts'o v = le16_to_cpu(es->s_want_extra_isize); 41439803387cSTheodore Ts'o if (v > max) { 41449803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 41459803387cSTheodore Ts'o "bad s_want_extra_isize: %d", v); 41469803387cSTheodore Ts'o goto failed_mount; 41479803387cSTheodore Ts'o } 41489803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 41499803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 41509803387cSTheodore Ts'o 41519803387cSTheodore Ts'o v = le16_to_cpu(es->s_min_extra_isize); 41529803387cSTheodore Ts'o if (v > max) { 41539803387cSTheodore Ts'o ext4_msg(sb, KERN_ERR, 41549803387cSTheodore Ts'o "bad s_min_extra_isize: %d", v); 41559803387cSTheodore Ts'o goto failed_mount; 41569803387cSTheodore Ts'o } 41579803387cSTheodore Ts'o if (sbi->s_want_extra_isize < v) 41589803387cSTheodore Ts'o sbi->s_want_extra_isize = v; 41599803387cSTheodore Ts'o } 41609803387cSTheodore Ts'o } 41619803387cSTheodore Ts'o 41625aee0f8aSTheodore Ts'o if (sbi->s_es->s_mount_opts[0]) { 41635aee0f8aSTheodore Ts'o char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts, 41645aee0f8aSTheodore Ts'o sizeof(sbi->s_es->s_mount_opts), 41655aee0f8aSTheodore Ts'o GFP_KERNEL); 41665aee0f8aSTheodore Ts'o if (!s_mount_opts) 41675aee0f8aSTheodore Ts'o goto failed_mount; 41685aee0f8aSTheodore Ts'o if (!parse_options(s_mount_opts, sb, &journal_devnum, 41695aee0f8aSTheodore Ts'o &journal_ioprio, 0)) { 41708b67f04aSTheodore Ts'o ext4_msg(sb, KERN_WARNING, 41718b67f04aSTheodore Ts'o "failed to parse options in superblock: %s", 41725aee0f8aSTheodore Ts'o s_mount_opts); 41735aee0f8aSTheodore Ts'o } 41745aee0f8aSTheodore Ts'o kfree(s_mount_opts); 41758b67f04aSTheodore Ts'o } 41765a916be1STheodore Ts'o sbi->s_def_mount_opt = sbi->s_mount_opt; 4177b3881f74STheodore Ts'o if (!parse_options((char *) data, sb, &journal_devnum, 4178661aa520SEric Sandeen &journal_ioprio, 0)) 4179ac27a0ecSDave Kleikamp goto failed_mount; 4180ac27a0ecSDave Kleikamp 4181c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4182c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) { 4183c83ad55eSGabriel Krisman Bertazi const struct ext4_sb_encodings *encoding_info; 4184c83ad55eSGabriel Krisman Bertazi struct unicode_map *encoding; 4185c83ad55eSGabriel Krisman Bertazi __u16 encoding_flags; 4186c83ad55eSGabriel Krisman Bertazi 4187c83ad55eSGabriel Krisman Bertazi if (ext4_has_feature_encrypt(sb)) { 4188c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4189c83ad55eSGabriel Krisman Bertazi "Can't mount with encoding and encryption"); 4190c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4191c83ad55eSGabriel Krisman Bertazi } 4192c83ad55eSGabriel Krisman Bertazi 4193c83ad55eSGabriel Krisman Bertazi if (ext4_sb_read_encoding(es, &encoding_info, 4194c83ad55eSGabriel Krisman Bertazi &encoding_flags)) { 4195c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4196c83ad55eSGabriel Krisman Bertazi "Encoding requested by superblock is unknown"); 4197c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4198c83ad55eSGabriel Krisman Bertazi } 4199c83ad55eSGabriel Krisman Bertazi 4200c83ad55eSGabriel Krisman Bertazi encoding = utf8_load(encoding_info->version); 4201c83ad55eSGabriel Krisman Bertazi if (IS_ERR(encoding)) { 4202c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_ERR, 4203c83ad55eSGabriel Krisman Bertazi "can't mount with superblock charset: %s-%s " 4204c83ad55eSGabriel Krisman Bertazi "not supported by the kernel. flags: 0x%x.", 4205c83ad55eSGabriel Krisman Bertazi encoding_info->name, encoding_info->version, 4206c83ad55eSGabriel Krisman Bertazi encoding_flags); 4207c83ad55eSGabriel Krisman Bertazi goto failed_mount; 4208c83ad55eSGabriel Krisman Bertazi } 4209c83ad55eSGabriel Krisman Bertazi ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: " 4210c83ad55eSGabriel Krisman Bertazi "%s-%s with flags 0x%hx", encoding_info->name, 4211c83ad55eSGabriel Krisman Bertazi encoding_info->version?:"\b", encoding_flags); 4212c83ad55eSGabriel Krisman Bertazi 4213c83ad55eSGabriel Krisman Bertazi sbi->s_encoding = encoding; 4214c83ad55eSGabriel Krisman Bertazi sbi->s_encoding_flags = encoding_flags; 4215c83ad55eSGabriel Krisman Bertazi } 4216c83ad55eSGabriel Krisman Bertazi #endif 4217c83ad55eSGabriel Krisman Bertazi 421856889787STheodore Ts'o if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 4219244adf64STheodore Ts'o printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n"); 4220781c036bSKaixu Xia /* can't mount with both data=journal and dioread_nolock. */ 4221244adf64STheodore Ts'o clear_opt(sb, DIOREAD_NOLOCK); 422256889787STheodore Ts'o if (test_opt2(sb, EXPLICIT_DELALLOC)) { 422356889787STheodore Ts'o ext4_msg(sb, KERN_ERR, "can't mount with " 422456889787STheodore Ts'o "both data=journal and delalloc"); 422556889787STheodore Ts'o goto failed_mount; 422656889787STheodore Ts'o } 4227fc626fe3SIra Weiny if (test_opt(sb, DAX_ALWAYS)) { 4228923ae0ffSRoss Zwisler ext4_msg(sb, KERN_ERR, "can't mount with " 4229923ae0ffSRoss Zwisler "both data=journal and dax"); 4230923ae0ffSRoss Zwisler goto failed_mount; 4231923ae0ffSRoss Zwisler } 423273b92a2aSSergey Karamov if (ext4_has_feature_encrypt(sb)) { 423373b92a2aSSergey Karamov ext4_msg(sb, KERN_WARNING, 423473b92a2aSSergey Karamov "encrypted files will use data=ordered " 423573b92a2aSSergey Karamov "instead of data journaling mode"); 423673b92a2aSSergey Karamov } 423756889787STheodore Ts'o if (test_opt(sb, DELALLOC)) 423856889787STheodore Ts'o clear_opt(sb, DELALLOC); 4239001e4a87STejun Heo } else { 4240001e4a87STejun Heo sb->s_iflags |= SB_I_CGROUPWB; 424156889787STheodore Ts'o } 424256889787STheodore Ts'o 42431751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 42441751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 4245ac27a0ecSDave Kleikamp 4246617ba13bSMingming Cao if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV && 4247e2b911c5SDarrick J. Wong (ext4_has_compat_features(sb) || 4248e2b911c5SDarrick J. Wong ext4_has_ro_compat_features(sb) || 4249e2b911c5SDarrick J. Wong ext4_has_incompat_features(sb))) 4250b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 4251b31e1552SEric Sandeen "feature flags set on rev 0 fs, " 4252b31e1552SEric Sandeen "running e2fsck is recommended"); 4253469108ffSTheodore Tso 4254ed3654ebSTheodore Ts'o if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) { 4255ed3654ebSTheodore Ts'o set_opt2(sb, HURD_COMPAT); 4256e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 4257ed3654ebSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4258ed3654ebSTheodore Ts'o "The Hurd can't support 64-bit file systems"); 4259ed3654ebSTheodore Ts'o goto failed_mount; 4260ed3654ebSTheodore Ts'o } 4261dec214d0STahsin Erdogan 4262dec214d0STahsin Erdogan /* 4263dec214d0STahsin Erdogan * ea_inode feature uses l_i_version field which is not 4264dec214d0STahsin Erdogan * available in HURD_COMPAT mode. 4265dec214d0STahsin Erdogan */ 4266dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4267dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4268dec214d0STahsin Erdogan "ea_inode feature is not supported for Hurd"); 4269dec214d0STahsin Erdogan goto failed_mount; 4270dec214d0STahsin Erdogan } 4271ed3654ebSTheodore Ts'o } 4272ed3654ebSTheodore Ts'o 42732035e776STheodore Ts'o if (IS_EXT2_SB(sb)) { 42742035e776STheodore Ts'o if (ext2_feature_set_ok(sb)) 42752035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext2 file system " 42762035e776STheodore Ts'o "using the ext4 subsystem"); 42772035e776STheodore Ts'o else { 42780d9366d6SEric Sandeen /* 42790d9366d6SEric Sandeen * If we're probing be silent, if this looks like 42800d9366d6SEric Sandeen * it's actually an ext[34] filesystem. 42810d9366d6SEric Sandeen */ 42820d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 42830d9366d6SEric Sandeen goto failed_mount; 42842035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due " 42852035e776STheodore Ts'o "to feature incompatibilities"); 42862035e776STheodore Ts'o goto failed_mount; 42872035e776STheodore Ts'o } 42882035e776STheodore Ts'o } 42892035e776STheodore Ts'o 42902035e776STheodore Ts'o if (IS_EXT3_SB(sb)) { 42912035e776STheodore Ts'o if (ext3_feature_set_ok(sb)) 42922035e776STheodore Ts'o ext4_msg(sb, KERN_INFO, "mounting ext3 file system " 42932035e776STheodore Ts'o "using the ext4 subsystem"); 42942035e776STheodore Ts'o else { 42950d9366d6SEric Sandeen /* 42960d9366d6SEric Sandeen * If we're probing be silent, if this looks like 42970d9366d6SEric Sandeen * it's actually an ext4 filesystem. 42980d9366d6SEric Sandeen */ 42990d9366d6SEric Sandeen if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb))) 43000d9366d6SEric Sandeen goto failed_mount; 43012035e776STheodore Ts'o ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due " 43022035e776STheodore Ts'o "to feature incompatibilities"); 43032035e776STheodore Ts'o goto failed_mount; 43042035e776STheodore Ts'o } 43052035e776STheodore Ts'o } 43062035e776STheodore Ts'o 4307469108ffSTheodore Tso /* 4308ac27a0ecSDave Kleikamp * Check feature flags regardless of the revision level, since we 4309ac27a0ecSDave Kleikamp * previously didn't change the revision level when setting the flags, 4310ac27a0ecSDave Kleikamp * so there is a chance incompat flags are set on a rev 0 filesystem. 4311ac27a0ecSDave Kleikamp */ 4312bc98a42cSDavid Howells if (!ext4_feature_set_ok(sb, (sb_rdonly(sb)))) 4313ac27a0ecSDave Kleikamp goto failed_mount; 4314a13fb1a4SEric Sandeen 43158cdf3372STheodore Ts'o if (le32_to_cpu(es->s_log_block_size) > 43168cdf3372STheodore Ts'o (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 43178cdf3372STheodore Ts'o ext4_msg(sb, KERN_ERR, 43188cdf3372STheodore Ts'o "Invalid log block size: %u", 43198cdf3372STheodore Ts'o le32_to_cpu(es->s_log_block_size)); 4320ac27a0ecSDave Kleikamp goto failed_mount; 4321ac27a0ecSDave Kleikamp } 4322bfe0a5f4STheodore Ts'o if (le32_to_cpu(es->s_log_cluster_size) > 4323bfe0a5f4STheodore Ts'o (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) { 4324bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4325bfe0a5f4STheodore Ts'o "Invalid log cluster size: %u", 4326bfe0a5f4STheodore Ts'o le32_to_cpu(es->s_log_cluster_size)); 4327bfe0a5f4STheodore Ts'o goto failed_mount; 4328bfe0a5f4STheodore Ts'o } 4329ac27a0ecSDave Kleikamp 43305b9554dcSTheodore Ts'o if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) { 43315b9554dcSTheodore Ts'o ext4_msg(sb, KERN_ERR, 43325b9554dcSTheodore Ts'o "Number of reserved GDT blocks insanely large: %d", 43335b9554dcSTheodore Ts'o le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks)); 43345b9554dcSTheodore Ts'o goto failed_mount; 43355b9554dcSTheodore Ts'o } 43365b9554dcSTheodore Ts'o 4337a8ab6d38SIra Weiny if (bdev_dax_supported(sb->s_bdev, blocksize)) 4338a8ab6d38SIra Weiny set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags); 4339a8ab6d38SIra Weiny 4340fc626fe3SIra Weiny if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) { 4341559db4c6SRoss Zwisler if (ext4_has_feature_inline_data(sb)) { 4342559db4c6SRoss Zwisler ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem" 4343559db4c6SRoss Zwisler " that may contain inline data"); 4344361d24d4SEric Sandeen goto failed_mount; 4345559db4c6SRoss Zwisler } 4346a8ab6d38SIra Weiny if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) { 434724f3478dSDan Williams ext4_msg(sb, KERN_ERR, 4348361d24d4SEric Sandeen "DAX unsupported by block device."); 4349361d24d4SEric Sandeen goto failed_mount; 435024f3478dSDan Williams } 4351923ae0ffSRoss Zwisler } 4352923ae0ffSRoss Zwisler 4353e2b911c5SDarrick J. Wong if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) { 43546ddb2447STheodore Ts'o ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d", 43556ddb2447STheodore Ts'o es->s_encryption_level); 43566ddb2447STheodore Ts'o goto failed_mount; 43576ddb2447STheodore Ts'o } 43586ddb2447STheodore Ts'o 4359ac27a0ecSDave Kleikamp if (sb->s_blocksize != blocksize) { 4360ce40733cSAneesh Kumar K.V /* Validate the filesystem blocksize */ 4361ce40733cSAneesh Kumar K.V if (!sb_set_blocksize(sb, blocksize)) { 4362b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "bad block size %d", 4363ce40733cSAneesh Kumar K.V blocksize); 4364ac27a0ecSDave Kleikamp goto failed_mount; 4365ac27a0ecSDave Kleikamp } 4366ac27a0ecSDave Kleikamp 4367ac27a0ecSDave Kleikamp brelse(bh); 436870bbb3e0SAndrew Morton logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE; 436970bbb3e0SAndrew Morton offset = do_div(logical_sb_block, blocksize); 43708394a6abSzhangyi (F) bh = ext4_sb_bread_unmovable(sb, logical_sb_block); 43718394a6abSzhangyi (F) if (IS_ERR(bh)) { 4372b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4373b31e1552SEric Sandeen "Can't read superblock on 2nd try"); 43748394a6abSzhangyi (F) ret = PTR_ERR(bh); 43758394a6abSzhangyi (F) bh = NULL; 4376ac27a0ecSDave Kleikamp goto failed_mount; 4377ac27a0ecSDave Kleikamp } 43782716b802STheodore Ts'o es = (struct ext4_super_block *)(bh->b_data + offset); 4379ac27a0ecSDave Kleikamp sbi->s_es = es; 4380617ba13bSMingming Cao if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) { 4381b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4382b31e1552SEric Sandeen "Magic mismatch, very weird!"); 4383ac27a0ecSDave Kleikamp goto failed_mount; 4384ac27a0ecSDave Kleikamp } 4385ac27a0ecSDave Kleikamp } 4386ac27a0ecSDave Kleikamp 4387e2b911c5SDarrick J. Wong has_huge_files = ext4_has_feature_huge_file(sb); 4388f287a1a5STheodore Ts'o sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits, 4389f287a1a5STheodore Ts'o has_huge_files); 4390f287a1a5STheodore Ts'o sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files); 4391ac27a0ecSDave Kleikamp 43920d1ee42fSAlexandre Ratchov sbi->s_desc_size = le16_to_cpu(es->s_desc_size); 4393e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb)) { 43948fadc143SAlexandre Ratchov if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT || 43950d1ee42fSAlexandre Ratchov sbi->s_desc_size > EXT4_MAX_DESC_SIZE || 4396d8ea6cf8Svignesh babu !is_power_of_2(sbi->s_desc_size)) { 4397b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4398b31e1552SEric Sandeen "unsupported descriptor size %lu", 43990d1ee42fSAlexandre Ratchov sbi->s_desc_size); 44000d1ee42fSAlexandre Ratchov goto failed_mount; 44010d1ee42fSAlexandre Ratchov } 44020d1ee42fSAlexandre Ratchov } else 44030d1ee42fSAlexandre Ratchov sbi->s_desc_size = EXT4_MIN_DESC_SIZE; 44040b8e58a1SAndreas Dilger 4405ac27a0ecSDave Kleikamp sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group); 4406ac27a0ecSDave Kleikamp sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group); 44070b8e58a1SAndreas Dilger 4408617ba13bSMingming Cao sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb); 4409ac27a0ecSDave Kleikamp if (sbi->s_inodes_per_block == 0) 4410617ba13bSMingming Cao goto cantfind_ext4; 4411cd6bb35bSTheodore Ts'o if (sbi->s_inodes_per_group < sbi->s_inodes_per_block || 4412cd6bb35bSTheodore Ts'o sbi->s_inodes_per_group > blocksize * 8) { 4413cd6bb35bSTheodore Ts'o ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n", 4414b9c538daSJosh Triplett sbi->s_inodes_per_group); 4415cd6bb35bSTheodore Ts'o goto failed_mount; 4416cd6bb35bSTheodore Ts'o } 4417ac27a0ecSDave Kleikamp sbi->s_itb_per_group = sbi->s_inodes_per_group / 4418ac27a0ecSDave Kleikamp sbi->s_inodes_per_block; 44190d1ee42fSAlexandre Ratchov sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb); 4420ac27a0ecSDave Kleikamp sbi->s_sbh = bh; 4421ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 4422e57aa839SFengguang Wu sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb)); 4423e57aa839SFengguang Wu sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb)); 44240b8e58a1SAndreas Dilger 4425ac27a0ecSDave Kleikamp for (i = 0; i < 4; i++) 4426ac27a0ecSDave Kleikamp sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); 4427ac27a0ecSDave Kleikamp sbi->s_def_hash_version = es->s_def_hash_version; 4428e2b911c5SDarrick J. Wong if (ext4_has_feature_dir_index(sb)) { 4429f99b2589STheodore Ts'o i = le32_to_cpu(es->s_flags); 4430f99b2589STheodore Ts'o if (i & EXT2_FLAGS_UNSIGNED_HASH) 4431f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4432f99b2589STheodore Ts'o else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) { 4433f99b2589STheodore Ts'o #ifdef __CHAR_UNSIGNED__ 4434bc98a42cSDavid Howells if (!sb_rdonly(sb)) 443523301410STheodore Ts'o es->s_flags |= 443623301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH); 4437f99b2589STheodore Ts'o sbi->s_hash_unsigned = 3; 4438f99b2589STheodore Ts'o #else 4439bc98a42cSDavid Howells if (!sb_rdonly(sb)) 444023301410STheodore Ts'o es->s_flags |= 444123301410STheodore Ts'o cpu_to_le32(EXT2_FLAGS_SIGNED_HASH); 4442f99b2589STheodore Ts'o #endif 4443f99b2589STheodore Ts'o } 444423301410STheodore Ts'o } 4445ac27a0ecSDave Kleikamp 4446281b5995STheodore Ts'o /* Handle clustersize */ 4447281b5995STheodore Ts'o clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size); 4448ef5fd681SKaixu Xia if (ext4_has_feature_bigalloc(sb)) { 4449281b5995STheodore Ts'o if (clustersize < blocksize) { 4450281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4451281b5995STheodore Ts'o "cluster size (%d) smaller than " 4452281b5995STheodore Ts'o "block size (%d)", clustersize, blocksize); 4453281b5995STheodore Ts'o goto failed_mount; 4454281b5995STheodore Ts'o } 4455281b5995STheodore Ts'o sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) - 4456281b5995STheodore Ts'o le32_to_cpu(es->s_log_block_size); 4457281b5995STheodore Ts'o sbi->s_clusters_per_group = 4458281b5995STheodore Ts'o le32_to_cpu(es->s_clusters_per_group); 4459281b5995STheodore Ts'o if (sbi->s_clusters_per_group > blocksize * 8) { 4460281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, 4461281b5995STheodore Ts'o "#clusters per group too big: %lu", 4462281b5995STheodore Ts'o sbi->s_clusters_per_group); 4463281b5995STheodore Ts'o goto failed_mount; 4464281b5995STheodore Ts'o } 4465281b5995STheodore Ts'o if (sbi->s_blocks_per_group != 4466281b5995STheodore Ts'o (sbi->s_clusters_per_group * (clustersize / blocksize))) { 4467281b5995STheodore Ts'o ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and " 4468281b5995STheodore Ts'o "clusters per group (%lu) inconsistent", 4469281b5995STheodore Ts'o sbi->s_blocks_per_group, 4470281b5995STheodore Ts'o sbi->s_clusters_per_group); 4471281b5995STheodore Ts'o goto failed_mount; 4472281b5995STheodore Ts'o } 4473281b5995STheodore Ts'o } else { 4474281b5995STheodore Ts'o if (clustersize != blocksize) { 4475bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_ERR, 4476bfe0a5f4STheodore Ts'o "fragment/cluster size (%d) != " 4477bfe0a5f4STheodore Ts'o "block size (%d)", clustersize, blocksize); 4478bfe0a5f4STheodore Ts'o goto failed_mount; 4479281b5995STheodore Ts'o } 4480ac27a0ecSDave Kleikamp if (sbi->s_blocks_per_group > blocksize * 8) { 4481b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4482b31e1552SEric Sandeen "#blocks per group too big: %lu", 4483ac27a0ecSDave Kleikamp sbi->s_blocks_per_group); 4484ac27a0ecSDave Kleikamp goto failed_mount; 4485ac27a0ecSDave Kleikamp } 4486281b5995STheodore Ts'o sbi->s_clusters_per_group = sbi->s_blocks_per_group; 4487281b5995STheodore Ts'o sbi->s_cluster_bits = 0; 4488281b5995STheodore Ts'o } 4489281b5995STheodore Ts'o sbi->s_cluster_ratio = clustersize / blocksize; 4490281b5995STheodore Ts'o 4491960fd856STheodore Ts'o /* Do we have standard group size of clustersize * 8 blocks ? */ 4492960fd856STheodore Ts'o if (sbi->s_blocks_per_group == clustersize << 3) 4493960fd856STheodore Ts'o set_opt2(sb, STD_GROUP_SIZE); 4494960fd856STheodore Ts'o 4495bf43d84bSEric Sandeen /* 4496bf43d84bSEric Sandeen * Test whether we have more sectors than will fit in sector_t, 4497bf43d84bSEric Sandeen * and whether the max offset is addressable by the page cache. 4498bf43d84bSEric Sandeen */ 44995a9ae68aSDarrick J. Wong err = generic_check_addressable(sb->s_blocksize_bits, 450030ca22c7SPatrick J. LoPresti ext4_blocks_count(es)); 45015a9ae68aSDarrick J. Wong if (err) { 4502b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "filesystem" 4503bf43d84bSEric Sandeen " too large to mount safely on this system"); 4504ac27a0ecSDave Kleikamp goto failed_mount; 4505ac27a0ecSDave Kleikamp } 4506ac27a0ecSDave Kleikamp 4507617ba13bSMingming Cao if (EXT4_BLOCKS_PER_GROUP(sb) == 0) 4508617ba13bSMingming Cao goto cantfind_ext4; 4509e7c95593SEric Sandeen 45100f2ddca6SFrom: Thiemo Nagel /* check blocks count against device size */ 45110f2ddca6SFrom: Thiemo Nagel blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits; 45120f2ddca6SFrom: Thiemo Nagel if (blocks_count && ext4_blocks_count(es) > blocks_count) { 4513b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu " 4514b31e1552SEric Sandeen "exceeds size of device (%llu blocks)", 45150f2ddca6SFrom: Thiemo Nagel ext4_blocks_count(es), blocks_count); 45160f2ddca6SFrom: Thiemo Nagel goto failed_mount; 45170f2ddca6SFrom: Thiemo Nagel } 45180f2ddca6SFrom: Thiemo Nagel 45194ec11028STheodore Ts'o /* 45204ec11028STheodore Ts'o * It makes no sense for the first data block to be beyond the end 45214ec11028STheodore Ts'o * of the filesystem. 45224ec11028STheodore Ts'o */ 45234ec11028STheodore Ts'o if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) { 4524b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4525b31e1552SEric Sandeen "block %u is beyond end of filesystem (%llu)", 4526e7c95593SEric Sandeen le32_to_cpu(es->s_first_data_block), 45274ec11028STheodore Ts'o ext4_blocks_count(es)); 4528e7c95593SEric Sandeen goto failed_mount; 4529e7c95593SEric Sandeen } 4530bfe0a5f4STheodore Ts'o if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) && 4531bfe0a5f4STheodore Ts'o (sbi->s_cluster_ratio == 1)) { 4532bfe0a5f4STheodore Ts'o ext4_msg(sb, KERN_WARNING, "bad geometry: first data " 4533bfe0a5f4STheodore Ts'o "block is 0 with a 1k block and cluster size"); 4534bfe0a5f4STheodore Ts'o goto failed_mount; 4535bfe0a5f4STheodore Ts'o } 4536bfe0a5f4STheodore Ts'o 4537bd81d8eeSLaurent Vivier blocks_count = (ext4_blocks_count(es) - 4538bd81d8eeSLaurent Vivier le32_to_cpu(es->s_first_data_block) + 4539bd81d8eeSLaurent Vivier EXT4_BLOCKS_PER_GROUP(sb) - 1); 4540bd81d8eeSLaurent Vivier do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb)); 45414ec11028STheodore Ts'o if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) { 4542df41460aSJosh Triplett ext4_msg(sb, KERN_WARNING, "groups count too large: %llu " 45434ec11028STheodore Ts'o "(block count %llu, first data block %u, " 4544df41460aSJosh Triplett "blocks per group %lu)", blocks_count, 45454ec11028STheodore Ts'o ext4_blocks_count(es), 45464ec11028STheodore Ts'o le32_to_cpu(es->s_first_data_block), 45474ec11028STheodore Ts'o EXT4_BLOCKS_PER_GROUP(sb)); 45484ec11028STheodore Ts'o goto failed_mount; 45494ec11028STheodore Ts'o } 4550bd81d8eeSLaurent Vivier sbi->s_groups_count = blocks_count; 4551fb0a387dSEric Sandeen sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count, 4552fb0a387dSEric Sandeen (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb))); 45539e463084STheodore Ts'o if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) != 45549e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count)) { 45559e463084STheodore Ts'o ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu", 45569e463084STheodore Ts'o le32_to_cpu(es->s_inodes_count), 45579e463084STheodore Ts'o ((u64)sbi->s_groups_count * sbi->s_inodes_per_group)); 45589e463084STheodore Ts'o ret = -EINVAL; 45599e463084STheodore Ts'o goto failed_mount; 45609e463084STheodore Ts'o } 4561617ba13bSMingming Cao db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) / 4562617ba13bSMingming Cao EXT4_DESC_PER_BLOCK(sb); 45633a4b77cdSEryu Guan if (ext4_has_feature_meta_bg(sb)) { 45642ba3e6e8STheodore Ts'o if (le32_to_cpu(es->s_first_meta_bg) > db_count) { 45653a4b77cdSEryu Guan ext4_msg(sb, KERN_WARNING, 45663a4b77cdSEryu Guan "first meta block group too large: %u " 45673a4b77cdSEryu Guan "(group descriptor block count %u)", 45683a4b77cdSEryu Guan le32_to_cpu(es->s_first_meta_bg), db_count); 45693a4b77cdSEryu Guan goto failed_mount; 45703a4b77cdSEryu Guan } 45713a4b77cdSEryu Guan } 45721d0c3924STheodore Ts'o rcu_assign_pointer(sbi->s_group_desc, 45731d0c3924STheodore Ts'o kvmalloc_array(db_count, 4574f18a5f21STheodore Ts'o sizeof(struct buffer_head *), 45751d0c3924STheodore Ts'o GFP_KERNEL)); 4576ac27a0ecSDave Kleikamp if (sbi->s_group_desc == NULL) { 4577b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "not enough memory"); 45782cde417dSTheodore Ts'o ret = -ENOMEM; 4579ac27a0ecSDave Kleikamp goto failed_mount; 4580ac27a0ecSDave Kleikamp } 4581ac27a0ecSDave Kleikamp 4582705895b6SPekka Enberg bgl_lock_init(sbi->s_blockgroup_lock); 4583ac27a0ecSDave Kleikamp 458485c8f176SAndrew Perepechko /* Pre-read the descriptors into the buffer cache */ 458585c8f176SAndrew Perepechko for (i = 0; i < db_count; i++) { 458685c8f176SAndrew Perepechko block = descriptor_loc(sb, logical_sb_block, i); 45875df1d412Szhangyi (F) ext4_sb_breadahead_unmovable(sb, block); 458885c8f176SAndrew Perepechko } 458985c8f176SAndrew Perepechko 4590ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) { 45911d0c3924STheodore Ts'o struct buffer_head *bh; 45921d0c3924STheodore Ts'o 459370bbb3e0SAndrew Morton block = descriptor_loc(sb, logical_sb_block, i); 45948394a6abSzhangyi (F) bh = ext4_sb_bread_unmovable(sb, block); 45958394a6abSzhangyi (F) if (IS_ERR(bh)) { 4596b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 4597b31e1552SEric Sandeen "can't read group descriptor %d", i); 4598ac27a0ecSDave Kleikamp db_count = i; 45998394a6abSzhangyi (F) ret = PTR_ERR(bh); 46008394a6abSzhangyi (F) bh = NULL; 4601ac27a0ecSDave Kleikamp goto failed_mount2; 4602ac27a0ecSDave Kleikamp } 46031d0c3924STheodore Ts'o rcu_read_lock(); 46041d0c3924STheodore Ts'o rcu_dereference(sbi->s_group_desc)[i] = bh; 46051d0c3924STheodore Ts'o rcu_read_unlock(); 4606ac27a0ecSDave Kleikamp } 460744de022cSTheodore Ts'o sbi->s_gdb_count = db_count; 4608829fa70dSTheodore Ts'o if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) { 4609b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 46106a797d27SDarrick J. Wong ret = -EFSCORRUPTED; 4611f9ae9cf5STheodore Ts'o goto failed_mount2; 4612ac27a0ecSDave Kleikamp } 4613772cb7c8SJose R. Santos 4614235699a8SKees Cook timer_setup(&sbi->s_err_report, print_daily_error_info, 0); 461504496411STao Ma 4616a75ae78fSDmitry Monakhov /* Register extent status tree shrinker */ 4617eb68d0e2SZheng Liu if (ext4_es_register_shrinker(sbi)) 4618ce7e010aSTheodore Ts'o goto failed_mount3; 4619ce7e010aSTheodore Ts'o 4620c9de560dSAlex Tomas sbi->s_stripe = ext4_get_stripe_size(sbi); 462167a5da56SZheng Liu sbi->s_extent_max_zeroout_kb = 32; 4622c9de560dSAlex Tomas 4623f9ae9cf5STheodore Ts'o /* 4624f9ae9cf5STheodore Ts'o * set up enough so that it can read an inode 4625f9ae9cf5STheodore Ts'o */ 4626f9ae9cf5STheodore Ts'o sb->s_op = &ext4_sops; 4627617ba13bSMingming Cao sb->s_export_op = &ext4_export_ops; 4628617ba13bSMingming Cao sb->s_xattr = ext4_xattr_handlers; 4629643fa961SChandan Rajendra #ifdef CONFIG_FS_ENCRYPTION 4630a7550b30SJaegeuk Kim sb->s_cop = &ext4_cryptops; 4631ffcc4182SEric Biggers #endif 4632c93d8f88SEric Biggers #ifdef CONFIG_FS_VERITY 4633c93d8f88SEric Biggers sb->s_vop = &ext4_verityops; 4634c93d8f88SEric Biggers #endif 4635ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 4636617ba13bSMingming Cao sb->dq_op = &ext4_quota_operations; 4637e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb)) 46381fa5efe3SJan Kara sb->s_qcop = &dquot_quotactl_sysfile_ops; 4639262b4662SJan Kara else 4640262b4662SJan Kara sb->s_qcop = &ext4_qctl_operations; 4641689c958cSLi Xi sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ; 4642ac27a0ecSDave Kleikamp #endif 464385787090SChristoph Hellwig memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid)); 4644f2fa2ffcSAneesh Kumar K.V 4645ac27a0ecSDave Kleikamp INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ 46463b9d4ed2STheodore Ts'o mutex_init(&sbi->s_orphan_lock); 4647ac27a0ecSDave Kleikamp 4648ac27a0ecSDave Kleikamp sb->s_root = NULL; 4649ac27a0ecSDave Kleikamp 4650ac27a0ecSDave Kleikamp needs_recovery = (es->s_last_orphan != 0 || 4651e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)); 4652ac27a0ecSDave Kleikamp 4653bc98a42cSDavid Howells if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) 4654c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block))) 465550460fe8SDarrick J. Wong goto failed_mount3a; 4656c5e06d10SJohann Lombardi 4657ac27a0ecSDave Kleikamp /* 4658ac27a0ecSDave Kleikamp * The first inode we look at is the journal inode. Don't try 4659ac27a0ecSDave Kleikamp * root first: it may be modified in the journal! 4660ac27a0ecSDave Kleikamp */ 4661e2b911c5SDarrick J. Wong if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) { 46624753d8a2STheodore Ts'o err = ext4_load_journal(sb, es, journal_devnum); 46634753d8a2STheodore Ts'o if (err) 466450460fe8SDarrick J. Wong goto failed_mount3a; 4665bc98a42cSDavid Howells } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) && 4666e2b911c5SDarrick J. Wong ext4_has_feature_journal_needs_recovery(sb)) { 4667b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "required journal recovery " 4668b31e1552SEric Sandeen "suppressed and not mounted read-only"); 4669744692dcSJiaying Zhang goto failed_mount_wq; 4670ac27a0ecSDave Kleikamp } else { 46711e381f60SDmitry Monakhov /* Nojournal mode, all journal mount options are illegal */ 46721e381f60SDmitry Monakhov if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) { 46731e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 46741e381f60SDmitry Monakhov "journal_checksum, fs mounted w/o journal"); 46751e381f60SDmitry Monakhov goto failed_mount_wq; 46761e381f60SDmitry Monakhov } 46771e381f60SDmitry Monakhov if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 46781e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 46791e381f60SDmitry Monakhov "journal_async_commit, fs mounted w/o journal"); 46801e381f60SDmitry Monakhov goto failed_mount_wq; 46811e381f60SDmitry Monakhov } 46821e381f60SDmitry Monakhov if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) { 46831e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 46841e381f60SDmitry Monakhov "commit=%lu, fs mounted w/o journal", 46851e381f60SDmitry Monakhov sbi->s_commit_interval / HZ); 46861e381f60SDmitry Monakhov goto failed_mount_wq; 46871e381f60SDmitry Monakhov } 46881e381f60SDmitry Monakhov if (EXT4_MOUNT_DATA_FLAGS & 46891e381f60SDmitry Monakhov (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) { 46901e381f60SDmitry Monakhov ext4_msg(sb, KERN_ERR, "can't mount with " 46911e381f60SDmitry Monakhov "data=, fs mounted w/o journal"); 46921e381f60SDmitry Monakhov goto failed_mount_wq; 46931e381f60SDmitry Monakhov } 469450b29d8fSDebabrata Banerjee sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM; 46951e381f60SDmitry Monakhov clear_opt(sb, JOURNAL_CHECKSUM); 4696fd8c37ecSTheodore Ts'o clear_opt(sb, DATA_FLAGS); 46970390131bSFrank Mayhar sbi->s_journal = NULL; 46980390131bSFrank Mayhar needs_recovery = 0; 46990390131bSFrank Mayhar goto no_journal; 4700ac27a0ecSDave Kleikamp } 4701ac27a0ecSDave Kleikamp 4702e2b911c5SDarrick J. Wong if (ext4_has_feature_64bit(sb) && 4703eb40a09cSJose R. Santos !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0, 4704eb40a09cSJose R. Santos JBD2_FEATURE_INCOMPAT_64BIT)) { 4705b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature"); 4706744692dcSJiaying Zhang goto failed_mount_wq; 4707eb40a09cSJose R. Santos } 4708eb40a09cSJose R. Santos 470925ed6e8aSDarrick J. Wong if (!set_journal_csum_feature_set(sb)) { 471025ed6e8aSDarrick J. Wong ext4_msg(sb, KERN_ERR, "Failed to set journal checksum " 471125ed6e8aSDarrick J. Wong "feature set"); 471225ed6e8aSDarrick J. Wong goto failed_mount_wq; 4713d4da6c9cSLinus Torvalds } 4714818d276cSGirish Shilamkar 4715ac27a0ecSDave Kleikamp /* We have now updated the journal if required, so we can 4716ac27a0ecSDave Kleikamp * validate the data journaling mode. */ 4717ac27a0ecSDave Kleikamp switch (test_opt(sb, DATA_FLAGS)) { 4718ac27a0ecSDave Kleikamp case 0: 4719ac27a0ecSDave Kleikamp /* No mode set, assume a default based on the journal 472063f57933SAndrew Morton * capabilities: ORDERED_DATA if the journal can 472163f57933SAndrew Morton * cope, else JOURNAL_DATA 472263f57933SAndrew Morton */ 4723dab291afSMingming Cao if (jbd2_journal_check_available_features 472427f394a7STyson Nottingham (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4725fd8c37ecSTheodore Ts'o set_opt(sb, ORDERED_DATA); 472627f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA; 472727f394a7STyson Nottingham } else { 4728fd8c37ecSTheodore Ts'o set_opt(sb, JOURNAL_DATA); 472927f394a7STyson Nottingham sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA; 473027f394a7STyson Nottingham } 4731ac27a0ecSDave Kleikamp break; 4732ac27a0ecSDave Kleikamp 4733617ba13bSMingming Cao case EXT4_MOUNT_ORDERED_DATA: 4734617ba13bSMingming Cao case EXT4_MOUNT_WRITEBACK_DATA: 4735dab291afSMingming Cao if (!jbd2_journal_check_available_features 4736dab291afSMingming Cao (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) { 4737b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Journal does not support " 4738b31e1552SEric Sandeen "requested data journaling mode"); 4739744692dcSJiaying Zhang goto failed_mount_wq; 4740ac27a0ecSDave Kleikamp } 4741ac27a0ecSDave Kleikamp default: 4742ac27a0ecSDave Kleikamp break; 4743ac27a0ecSDave Kleikamp } 4744ab04df78SJan Kara 4745ab04df78SJan Kara if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA && 4746ab04df78SJan Kara test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 4747ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 4748ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 4749ab04df78SJan Kara goto failed_mount_wq; 4750ab04df78SJan Kara } 4751ab04df78SJan Kara 4752b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 4753ac27a0ecSDave Kleikamp 475418aadd47SBobi Jam sbi->s_journal->j_commit_callback = ext4_journal_commit_callback; 4755*342af94eSMauricio Faria de Oliveira sbi->s_journal->j_submit_inode_data_buffers = 4756*342af94eSMauricio Faria de Oliveira jbd2_journal_submit_inode_data_buffers; 4757*342af94eSMauricio Faria de Oliveira sbi->s_journal->j_finish_inode_data_buffers = 4758*342af94eSMauricio Faria de Oliveira jbd2_journal_finish_inode_data_buffers; 475918aadd47SBobi Jam 4760ce7e010aSTheodore Ts'o no_journal: 4761cdb7ee4cSTahsin Erdogan if (!test_opt(sb, NO_MBCACHE)) { 476247387409STahsin Erdogan sbi->s_ea_block_cache = ext4_xattr_create_cache(); 476347387409STahsin Erdogan if (!sbi->s_ea_block_cache) { 4764cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, 4765cdb7ee4cSTahsin Erdogan "Failed to create ea_block_cache"); 47669c191f70ST Makphaibulchoke goto failed_mount_wq; 47679c191f70ST Makphaibulchoke } 47689c191f70ST Makphaibulchoke 4769dec214d0STahsin Erdogan if (ext4_has_feature_ea_inode(sb)) { 4770dec214d0STahsin Erdogan sbi->s_ea_inode_cache = ext4_xattr_create_cache(); 4771dec214d0STahsin Erdogan if (!sbi->s_ea_inode_cache) { 4772dec214d0STahsin Erdogan ext4_msg(sb, KERN_ERR, 4773dec214d0STahsin Erdogan "Failed to create ea_inode_cache"); 4774dec214d0STahsin Erdogan goto failed_mount_wq; 4775dec214d0STahsin Erdogan } 4776dec214d0STahsin Erdogan } 4777cdb7ee4cSTahsin Erdogan } 4778dec214d0STahsin Erdogan 4779c93d8f88SEric Biggers if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) { 4780c93d8f88SEric Biggers ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity"); 4781c93d8f88SEric Biggers goto failed_mount_wq; 4782c93d8f88SEric Biggers } 4783c93d8f88SEric Biggers 4784bc98a42cSDavid Howells if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) && 4785e2b911c5SDarrick J. Wong !ext4_has_feature_encrypt(sb)) { 4786e2b911c5SDarrick J. Wong ext4_set_feature_encrypt(sb); 47876ddb2447STheodore Ts'o ext4_commit_super(sb, 1); 47886ddb2447STheodore Ts'o } 47896ddb2447STheodore Ts'o 4790fd89d5f2STejun Heo /* 4791952fc18eSTheodore Ts'o * Get the # of file system overhead blocks from the 4792952fc18eSTheodore Ts'o * superblock if present. 4793952fc18eSTheodore Ts'o */ 4794952fc18eSTheodore Ts'o if (es->s_overhead_clusters) 4795952fc18eSTheodore Ts'o sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters); 4796952fc18eSTheodore Ts'o else { 479707aa2ea1SLukas Czerner err = ext4_calculate_overhead(sb); 479807aa2ea1SLukas Czerner if (err) 4799952fc18eSTheodore Ts'o goto failed_mount_wq; 4800952fc18eSTheodore Ts'o } 4801952fc18eSTheodore Ts'o 4802952fc18eSTheodore Ts'o /* 4803fd89d5f2STejun Heo * The maximum number of concurrent works can be high and 4804fd89d5f2STejun Heo * concurrency isn't really necessary. Limit it to 1. 4805fd89d5f2STejun Heo */ 48062e8fa54eSJan Kara EXT4_SB(sb)->rsv_conversion_wq = 48072e8fa54eSJan Kara alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1); 48082e8fa54eSJan Kara if (!EXT4_SB(sb)->rsv_conversion_wq) { 48092e8fa54eSJan Kara printk(KERN_ERR "EXT4-fs: failed to create workqueue\n"); 481007aa2ea1SLukas Czerner ret = -ENOMEM; 48112e8fa54eSJan Kara goto failed_mount4; 48122e8fa54eSJan Kara } 48132e8fa54eSJan Kara 4814ac27a0ecSDave Kleikamp /* 4815dab291afSMingming Cao * The jbd2_journal_load will have done any necessary log recovery, 4816ac27a0ecSDave Kleikamp * so we can safely mount the rest of the filesystem now. 4817ac27a0ecSDave Kleikamp */ 4818ac27a0ecSDave Kleikamp 48198a363970STheodore Ts'o root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL); 48201d1fe1eeSDavid Howells if (IS_ERR(root)) { 4821b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root inode failed"); 48221d1fe1eeSDavid Howells ret = PTR_ERR(root); 482332a9bb57SManish Katiyar root = NULL; 4824ac27a0ecSDave Kleikamp goto failed_mount4; 4825ac27a0ecSDave Kleikamp } 4826ac27a0ecSDave Kleikamp if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) { 4827b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck"); 482894bf608aSAl Viro iput(root); 4829ac27a0ecSDave Kleikamp goto failed_mount4; 4830ac27a0ecSDave Kleikamp } 4831b886ee3eSGabriel Krisman Bertazi 4832b886ee3eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 4833b886ee3eSGabriel Krisman Bertazi if (sbi->s_encoding) 4834b886ee3eSGabriel Krisman Bertazi sb->s_d_op = &ext4_dentry_ops; 4835b886ee3eSGabriel Krisman Bertazi #endif 4836b886ee3eSGabriel Krisman Bertazi 483748fde701SAl Viro sb->s_root = d_make_root(root); 48381d1fe1eeSDavid Howells if (!sb->s_root) { 4839b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "get root dentry failed"); 48401d1fe1eeSDavid Howells ret = -ENOMEM; 48411d1fe1eeSDavid Howells goto failed_mount4; 48421d1fe1eeSDavid Howells } 4843ac27a0ecSDave Kleikamp 4844c89128a0SJaegeuk Kim ret = ext4_setup_super(sb, es, sb_rdonly(sb)); 4845c89128a0SJaegeuk Kim if (ret == -EROFS) { 48461751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 4847c89128a0SJaegeuk Kim ret = 0; 4848c89128a0SJaegeuk Kim } else if (ret) 4849c89128a0SJaegeuk Kim goto failed_mount4a; 4850ef7f3835SKalpak Shah 4851b5799018STheodore Ts'o ext4_set_resv_clusters(sb); 485227dd4385SLukas Czerner 48530f5bde1dSJan Kara if (test_opt(sb, BLOCK_VALIDITY)) { 48546fd058f7STheodore Ts'o err = ext4_setup_system_zone(sb); 48556fd058f7STheodore Ts'o if (err) { 4856b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to initialize system " 4857fbe845ddSCurt Wohlgemuth "zone (%d)", err); 4858f9ae9cf5STheodore Ts'o goto failed_mount4a; 4859f9ae9cf5STheodore Ts'o } 48600f5bde1dSJan Kara } 4861f9ae9cf5STheodore Ts'o 4862f9ae9cf5STheodore Ts'o ext4_ext_init(sb); 4863f9ae9cf5STheodore Ts'o err = ext4_mb_init(sb); 4864f9ae9cf5STheodore Ts'o if (err) { 4865f9ae9cf5STheodore Ts'o ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)", 4866f9ae9cf5STheodore Ts'o err); 4867dcf2d804STao Ma goto failed_mount5; 4868c2774d84SAneesh Kumar K.V } 4869c2774d84SAneesh Kumar K.V 4870d5e03cbbSTheodore Ts'o block = ext4_count_free_clusters(sb); 4871d5e03cbbSTheodore Ts'o ext4_free_blocks_count_set(sbi->s_es, 4872d5e03cbbSTheodore Ts'o EXT4_C2B(sbi, block)); 48734274f516STheodore Ts'o ext4_superblock_csum_set(sb); 4874908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeclusters_counter, block, 4875908c7f19STejun Heo GFP_KERNEL); 4876d5e03cbbSTheodore Ts'o if (!err) { 4877d5e03cbbSTheodore Ts'o unsigned long freei = ext4_count_free_inodes(sb); 4878d5e03cbbSTheodore Ts'o sbi->s_es->s_free_inodes_count = cpu_to_le32(freei); 48794274f516STheodore Ts'o ext4_superblock_csum_set(sb); 4880908c7f19STejun Heo err = percpu_counter_init(&sbi->s_freeinodes_counter, freei, 4881908c7f19STejun Heo GFP_KERNEL); 4882d5e03cbbSTheodore Ts'o } 4883d5e03cbbSTheodore Ts'o if (!err) 4884d5e03cbbSTheodore Ts'o err = percpu_counter_init(&sbi->s_dirs_counter, 4885908c7f19STejun Heo ext4_count_dirs(sb), GFP_KERNEL); 4886d5e03cbbSTheodore Ts'o if (!err) 4887908c7f19STejun Heo err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0, 4888908c7f19STejun Heo GFP_KERNEL); 4889c8585c6fSDaeho Jeong if (!err) 4890bbd55937SEric Biggers err = percpu_init_rwsem(&sbi->s_writepages_rwsem); 4891c8585c6fSDaeho Jeong 4892d5e03cbbSTheodore Ts'o if (err) { 4893d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, "insufficient memory"); 4894d5e03cbbSTheodore Ts'o goto failed_mount6; 4895d5e03cbbSTheodore Ts'o } 4896d5e03cbbSTheodore Ts'o 4897e2b911c5SDarrick J. Wong if (ext4_has_feature_flex_bg(sb)) 4898d5e03cbbSTheodore Ts'o if (!ext4_fill_flex_info(sb)) { 4899d5e03cbbSTheodore Ts'o ext4_msg(sb, KERN_ERR, 4900d5e03cbbSTheodore Ts'o "unable to initialize " 4901d5e03cbbSTheodore Ts'o "flex_bg meta info!"); 4902d5e03cbbSTheodore Ts'o goto failed_mount6; 4903d5e03cbbSTheodore Ts'o } 4904d5e03cbbSTheodore Ts'o 4905bfff6873SLukas Czerner err = ext4_register_li_request(sb, first_not_zeroed); 4906bfff6873SLukas Czerner if (err) 4907dcf2d804STao Ma goto failed_mount6; 4908bfff6873SLukas Czerner 4909b5799018STheodore Ts'o err = ext4_register_sysfs(sb); 4910dcf2d804STao Ma if (err) 4911dcf2d804STao Ma goto failed_mount7; 49123197ebdbSTheodore Ts'o 49139b2ff357SJan Kara #ifdef CONFIG_QUOTA 49149b2ff357SJan Kara /* Enable quota usage during mount. */ 4915bc98a42cSDavid Howells if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) { 49169b2ff357SJan Kara err = ext4_enable_quotas(sb); 49179b2ff357SJan Kara if (err) 49189b2ff357SJan Kara goto failed_mount8; 49199b2ff357SJan Kara } 49209b2ff357SJan Kara #endif /* CONFIG_QUOTA */ 49219b2ff357SJan Kara 4922bc71726cSzhangyi (F) /* 4923bc71726cSzhangyi (F) * Save the original bdev mapping's wb_err value which could be 4924bc71726cSzhangyi (F) * used to detect the metadata async write error. 4925bc71726cSzhangyi (F) */ 4926bc71726cSzhangyi (F) spin_lock_init(&sbi->s_bdev_wb_lock); 4927bc71726cSzhangyi (F) errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err, 4928bc71726cSzhangyi (F) &sbi->s_bdev_wb_err); 4929bc71726cSzhangyi (F) sb->s_bdev->bd_super = sb; 4930617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS; 4931617ba13bSMingming Cao ext4_orphan_cleanup(sb, es); 4932617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS; 49330390131bSFrank Mayhar if (needs_recovery) { 4934b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "recovery complete"); 493511215630SJan Kara err = ext4_mark_recovery_complete(sb, es); 493611215630SJan Kara if (err) 493711215630SJan Kara goto failed_mount8; 49380390131bSFrank Mayhar } 49390390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal) { 49400390131bSFrank Mayhar if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) 49410390131bSFrank Mayhar descr = " journalled data mode"; 49420390131bSFrank Mayhar else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) 49430390131bSFrank Mayhar descr = " ordered data mode"; 49440390131bSFrank Mayhar else 49450390131bSFrank Mayhar descr = " writeback data mode"; 49460390131bSFrank Mayhar } else 49470390131bSFrank Mayhar descr = "out journal"; 49480390131bSFrank Mayhar 494979add3a3SLukas Czerner if (test_opt(sb, DISCARD)) { 495079add3a3SLukas Czerner struct request_queue *q = bdev_get_queue(sb->s_bdev); 495179add3a3SLukas Czerner if (!blk_queue_discard(q)) 495279add3a3SLukas Czerner ext4_msg(sb, KERN_WARNING, 495379add3a3SLukas Czerner "mounting with \"discard\" option, but " 495479add3a3SLukas Czerner "the device does not support discard"); 495579add3a3SLukas Czerner } 495679add3a3SLukas Czerner 4957e294a537STheodore Ts'o if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount")) 4958d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 49595aee0f8aSTheodore Ts'o "Opts: %.*s%s%s", descr, 49605aee0f8aSTheodore Ts'o (int) sizeof(sbi->s_es->s_mount_opts), 49615aee0f8aSTheodore Ts'o sbi->s_es->s_mount_opts, 49628b67f04aSTheodore Ts'o *sbi->s_es->s_mount_opts ? "; " : "", orig_data); 4963ac27a0ecSDave Kleikamp 496466e61a9eSTheodore Ts'o if (es->s_error_count) 496566e61a9eSTheodore Ts'o mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */ 4966ac27a0ecSDave Kleikamp 4967efbed4dcSTheodore Ts'o /* Enable message ratelimiting. Default is 10 messages per 5 secs. */ 4968efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10); 4969efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10); 4970efbed4dcSTheodore Ts'o ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10); 49711cf006edSDmitry Monakhov atomic_set(&sbi->s_warning_count, 0); 49721cf006edSDmitry Monakhov atomic_set(&sbi->s_msg_count, 0); 4973efbed4dcSTheodore Ts'o 4974d4c402d9SCurt Wohlgemuth kfree(orig_data); 4975ac27a0ecSDave Kleikamp return 0; 4976ac27a0ecSDave Kleikamp 4977617ba13bSMingming Cao cantfind_ext4: 4978ac27a0ecSDave Kleikamp if (!silent) 4979b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem"); 4980ac27a0ecSDave Kleikamp goto failed_mount; 4981ac27a0ecSDave Kleikamp 498272ba7450STheodore Ts'o failed_mount8: 4983ebd173beSTheodore Ts'o ext4_unregister_sysfs(sb); 4984cb8d53d2SEric Biggers kobject_put(&sbi->s_kobj); 4985dcf2d804STao Ma failed_mount7: 4986dcf2d804STao Ma ext4_unregister_li_request(sb); 4987dcf2d804STao Ma failed_mount6: 4988f9ae9cf5STheodore Ts'o ext4_mb_release(sb); 49897c990728SSuraj Jitindar Singh rcu_read_lock(); 49907c990728SSuraj Jitindar Singh flex_groups = rcu_dereference(sbi->s_flex_groups); 49917c990728SSuraj Jitindar Singh if (flex_groups) { 49927c990728SSuraj Jitindar Singh for (i = 0; i < sbi->s_flex_groups_allocated; i++) 49937c990728SSuraj Jitindar Singh kvfree(flex_groups[i]); 49947c990728SSuraj Jitindar Singh kvfree(flex_groups); 49957c990728SSuraj Jitindar Singh } 49967c990728SSuraj Jitindar Singh rcu_read_unlock(); 4997d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeclusters_counter); 4998d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_freeinodes_counter); 4999d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirs_counter); 5000d5e03cbbSTheodore Ts'o percpu_counter_destroy(&sbi->s_dirtyclusters_counter); 5001bbd55937SEric Biggers percpu_free_rwsem(&sbi->s_writepages_rwsem); 500200764937SAzat Khuzhin failed_mount5: 5003f9ae9cf5STheodore Ts'o ext4_ext_release(sb); 5004f9ae9cf5STheodore Ts'o ext4_release_system_zone(sb); 5005f9ae9cf5STheodore Ts'o failed_mount4a: 500694bf608aSAl Viro dput(sb->s_root); 500732a9bb57SManish Katiyar sb->s_root = NULL; 500894bf608aSAl Viro failed_mount4: 5009b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "mount failed"); 50102e8fa54eSJan Kara if (EXT4_SB(sb)->rsv_conversion_wq) 50112e8fa54eSJan Kara destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq); 50124c0425ffSMingming Cao failed_mount_wq: 5013dec214d0STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_inode_cache); 5014dec214d0STahsin Erdogan sbi->s_ea_inode_cache = NULL; 501550c15df6SChengguang Xu 501647387409STahsin Erdogan ext4_xattr_destroy_cache(sbi->s_ea_block_cache); 501747387409STahsin Erdogan sbi->s_ea_block_cache = NULL; 501850c15df6SChengguang Xu 50190390131bSFrank Mayhar if (sbi->s_journal) { 5020dab291afSMingming Cao jbd2_journal_destroy(sbi->s_journal); 502147b4a50bSJan Kara sbi->s_journal = NULL; 50220390131bSFrank Mayhar } 502350460fe8SDarrick J. Wong failed_mount3a: 5024d3922a77SZheng Liu ext4_es_unregister_shrinker(sbi); 5025eb68d0e2SZheng Liu failed_mount3: 50269105bb14SAl Viro del_timer_sync(&sbi->s_err_report); 5027c5e06d10SJohann Lombardi if (sbi->s_mmp_tsk) 5028c5e06d10SJohann Lombardi kthread_stop(sbi->s_mmp_tsk); 5029ac27a0ecSDave Kleikamp failed_mount2: 50301d0c3924STheodore Ts'o rcu_read_lock(); 50311d0c3924STheodore Ts'o group_desc = rcu_dereference(sbi->s_group_desc); 5032ac27a0ecSDave Kleikamp for (i = 0; i < db_count; i++) 50331d0c3924STheodore Ts'o brelse(group_desc[i]); 50341d0c3924STheodore Ts'o kvfree(group_desc); 50351d0c3924STheodore Ts'o rcu_read_unlock(); 5036ac27a0ecSDave Kleikamp failed_mount: 50370441984aSDarrick J. Wong if (sbi->s_chksum_driver) 50380441984aSDarrick J. Wong crypto_free_shash(sbi->s_chksum_driver); 5039c83ad55eSGabriel Krisman Bertazi 5040c83ad55eSGabriel Krisman Bertazi #ifdef CONFIG_UNICODE 5041c83ad55eSGabriel Krisman Bertazi utf8_unload(sbi->s_encoding); 5042c83ad55eSGabriel Krisman Bertazi #endif 5043c83ad55eSGabriel Krisman Bertazi 5044ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5045a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 50460ba33facSTheodore Ts'o kfree(get_qf_name(sb, sbi, i)); 5047ac27a0ecSDave Kleikamp #endif 5048ed318a6cSEric Biggers fscrypt_free_dummy_context(&sbi->s_dummy_enc_ctx); 5049617ba13bSMingming Cao ext4_blkdev_remove(sbi); 5050ac27a0ecSDave Kleikamp brelse(bh); 5051ac27a0ecSDave Kleikamp out_fail: 5052ac27a0ecSDave Kleikamp sb->s_fs_info = NULL; 5053f6830165SManish Katiyar kfree(sbi->s_blockgroup_lock); 50545aee0f8aSTheodore Ts'o out_free_base: 5055ac27a0ecSDave Kleikamp kfree(sbi); 5056d4c402d9SCurt Wohlgemuth kfree(orig_data); 50575e405595SDan Williams fs_put_dax(dax_dev); 505807aa2ea1SLukas Czerner return err ? err : ret; 5059ac27a0ecSDave Kleikamp } 5060ac27a0ecSDave Kleikamp 5061ac27a0ecSDave Kleikamp /* 5062ac27a0ecSDave Kleikamp * Setup any per-fs journal parameters now. We'll do this both on 5063ac27a0ecSDave Kleikamp * initial mount, once the journal has been initialised but before we've 5064ac27a0ecSDave Kleikamp * done any recovery; and again on any subsequent remount. 5065ac27a0ecSDave Kleikamp */ 5066617ba13bSMingming Cao static void ext4_init_journal_params(struct super_block *sb, journal_t *journal) 5067ac27a0ecSDave Kleikamp { 5068617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5069ac27a0ecSDave Kleikamp 5070ac27a0ecSDave Kleikamp journal->j_commit_interval = sbi->s_commit_interval; 507130773840STheodore Ts'o journal->j_min_batch_time = sbi->s_min_batch_time; 507230773840STheodore Ts'o journal->j_max_batch_time = sbi->s_max_batch_time; 5073ac27a0ecSDave Kleikamp 5074a931da6aSTheodore Ts'o write_lock(&journal->j_state_lock); 5075ac27a0ecSDave Kleikamp if (test_opt(sb, BARRIER)) 5076dab291afSMingming Cao journal->j_flags |= JBD2_BARRIER; 5077ac27a0ecSDave Kleikamp else 5078dab291afSMingming Cao journal->j_flags &= ~JBD2_BARRIER; 50795bf5683aSHidehiro Kawai if (test_opt(sb, DATA_ERR_ABORT)) 50805bf5683aSHidehiro Kawai journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 50815bf5683aSHidehiro Kawai else 50825bf5683aSHidehiro Kawai journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 5083a931da6aSTheodore Ts'o write_unlock(&journal->j_state_lock); 5084ac27a0ecSDave Kleikamp } 5085ac27a0ecSDave Kleikamp 5086c6cb7e77SEric Whitney static struct inode *ext4_get_journal_inode(struct super_block *sb, 5087ac27a0ecSDave Kleikamp unsigned int journal_inum) 5088ac27a0ecSDave Kleikamp { 5089ac27a0ecSDave Kleikamp struct inode *journal_inode; 5090ac27a0ecSDave Kleikamp 5091c6cb7e77SEric Whitney /* 5092c6cb7e77SEric Whitney * Test for the existence of a valid inode on disk. Bad things 5093c6cb7e77SEric Whitney * happen if we iget() an unused inode, as the subsequent iput() 5094c6cb7e77SEric Whitney * will try to delete it. 5095c6cb7e77SEric Whitney */ 50968a363970STheodore Ts'o journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL); 50971d1fe1eeSDavid Howells if (IS_ERR(journal_inode)) { 5098b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "no journal found"); 5099ac27a0ecSDave Kleikamp return NULL; 5100ac27a0ecSDave Kleikamp } 5101ac27a0ecSDave Kleikamp if (!journal_inode->i_nlink) { 5102ac27a0ecSDave Kleikamp make_bad_inode(journal_inode); 5103ac27a0ecSDave Kleikamp iput(journal_inode); 5104b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal inode is deleted"); 5105ac27a0ecSDave Kleikamp return NULL; 5106ac27a0ecSDave Kleikamp } 5107ac27a0ecSDave Kleikamp 5108e5f8eab8STheodore Ts'o jbd_debug(2, "Journal inode found at %p: %lld bytes\n", 5109ac27a0ecSDave Kleikamp journal_inode, journal_inode->i_size); 51101d1fe1eeSDavid Howells if (!S_ISREG(journal_inode->i_mode)) { 5111b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "invalid journal inode"); 5112ac27a0ecSDave Kleikamp iput(journal_inode); 5113ac27a0ecSDave Kleikamp return NULL; 5114ac27a0ecSDave Kleikamp } 5115c6cb7e77SEric Whitney return journal_inode; 5116c6cb7e77SEric Whitney } 5117c6cb7e77SEric Whitney 5118c6cb7e77SEric Whitney static journal_t *ext4_get_journal(struct super_block *sb, 5119c6cb7e77SEric Whitney unsigned int journal_inum) 5120c6cb7e77SEric Whitney { 5121c6cb7e77SEric Whitney struct inode *journal_inode; 5122c6cb7e77SEric Whitney journal_t *journal; 5123c6cb7e77SEric Whitney 512411215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 512511215630SJan Kara return NULL; 5126c6cb7e77SEric Whitney 5127c6cb7e77SEric Whitney journal_inode = ext4_get_journal_inode(sb, journal_inum); 5128c6cb7e77SEric Whitney if (!journal_inode) 5129c6cb7e77SEric Whitney return NULL; 5130ac27a0ecSDave Kleikamp 5131dab291afSMingming Cao journal = jbd2_journal_init_inode(journal_inode); 5132ac27a0ecSDave Kleikamp if (!journal) { 5133b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "Could not load journal inode"); 5134ac27a0ecSDave Kleikamp iput(journal_inode); 5135ac27a0ecSDave Kleikamp return NULL; 5136ac27a0ecSDave Kleikamp } 5137ac27a0ecSDave Kleikamp journal->j_private = sb; 5138617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5139ac27a0ecSDave Kleikamp return journal; 5140ac27a0ecSDave Kleikamp } 5141ac27a0ecSDave Kleikamp 5142617ba13bSMingming Cao static journal_t *ext4_get_dev_journal(struct super_block *sb, 5143ac27a0ecSDave Kleikamp dev_t j_dev) 5144ac27a0ecSDave Kleikamp { 5145ac27a0ecSDave Kleikamp struct buffer_head *bh; 5146ac27a0ecSDave Kleikamp journal_t *journal; 5147617ba13bSMingming Cao ext4_fsblk_t start; 5148617ba13bSMingming Cao ext4_fsblk_t len; 5149ac27a0ecSDave Kleikamp int hblock, blocksize; 5150617ba13bSMingming Cao ext4_fsblk_t sb_block; 5151ac27a0ecSDave Kleikamp unsigned long offset; 5152617ba13bSMingming Cao struct ext4_super_block *es; 5153ac27a0ecSDave Kleikamp struct block_device *bdev; 5154ac27a0ecSDave Kleikamp 515511215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 515611215630SJan Kara return NULL; 51570390131bSFrank Mayhar 5158b31e1552SEric Sandeen bdev = ext4_blkdev_get(j_dev, sb); 5159ac27a0ecSDave Kleikamp if (bdev == NULL) 5160ac27a0ecSDave Kleikamp return NULL; 5161ac27a0ecSDave Kleikamp 5162ac27a0ecSDave Kleikamp blocksize = sb->s_blocksize; 5163e1defc4fSMartin K. Petersen hblock = bdev_logical_block_size(bdev); 5164ac27a0ecSDave Kleikamp if (blocksize < hblock) { 5165b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5166b31e1552SEric Sandeen "blocksize too small for journal device"); 5167ac27a0ecSDave Kleikamp goto out_bdev; 5168ac27a0ecSDave Kleikamp } 5169ac27a0ecSDave Kleikamp 5170617ba13bSMingming Cao sb_block = EXT4_MIN_BLOCK_SIZE / blocksize; 5171617ba13bSMingming Cao offset = EXT4_MIN_BLOCK_SIZE % blocksize; 5172ac27a0ecSDave Kleikamp set_blocksize(bdev, blocksize); 5173ac27a0ecSDave Kleikamp if (!(bh = __bread(bdev, sb_block, blocksize))) { 5174b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "couldn't read superblock of " 5175b31e1552SEric Sandeen "external journal"); 5176ac27a0ecSDave Kleikamp goto out_bdev; 5177ac27a0ecSDave Kleikamp } 5178ac27a0ecSDave Kleikamp 51792716b802STheodore Ts'o es = (struct ext4_super_block *) (bh->b_data + offset); 5180617ba13bSMingming Cao if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) || 5181ac27a0ecSDave Kleikamp !(le32_to_cpu(es->s_feature_incompat) & 5182617ba13bSMingming Cao EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) { 5183b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "external journal has " 5184b31e1552SEric Sandeen "bad superblock"); 5185ac27a0ecSDave Kleikamp brelse(bh); 5186ac27a0ecSDave Kleikamp goto out_bdev; 5187ac27a0ecSDave Kleikamp } 5188ac27a0ecSDave Kleikamp 5189df4763beSDarrick J. Wong if ((le32_to_cpu(es->s_feature_ro_compat) & 5190df4763beSDarrick J. Wong EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) && 5191df4763beSDarrick J. Wong es->s_checksum != ext4_superblock_csum(sb, es)) { 5192df4763beSDarrick J. Wong ext4_msg(sb, KERN_ERR, "external journal has " 5193df4763beSDarrick J. Wong "corrupt superblock"); 5194df4763beSDarrick J. Wong brelse(bh); 5195df4763beSDarrick J. Wong goto out_bdev; 5196df4763beSDarrick J. Wong } 5197df4763beSDarrick J. Wong 5198617ba13bSMingming Cao if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) { 5199b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "journal UUID does not match"); 5200ac27a0ecSDave Kleikamp brelse(bh); 5201ac27a0ecSDave Kleikamp goto out_bdev; 5202ac27a0ecSDave Kleikamp } 5203ac27a0ecSDave Kleikamp 5204bd81d8eeSLaurent Vivier len = ext4_blocks_count(es); 5205ac27a0ecSDave Kleikamp start = sb_block + 1; 5206ac27a0ecSDave Kleikamp brelse(bh); /* we're done with the superblock */ 5207ac27a0ecSDave Kleikamp 5208dab291afSMingming Cao journal = jbd2_journal_init_dev(bdev, sb->s_bdev, 5209ac27a0ecSDave Kleikamp start, len, blocksize); 5210ac27a0ecSDave Kleikamp if (!journal) { 5211b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "failed to create device journal"); 5212ac27a0ecSDave Kleikamp goto out_bdev; 5213ac27a0ecSDave Kleikamp } 5214ac27a0ecSDave Kleikamp journal->j_private = sb; 52152d069c08Szhangyi (F) if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) { 5216b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error on journal device"); 5217ac27a0ecSDave Kleikamp goto out_journal; 5218ac27a0ecSDave Kleikamp } 5219ac27a0ecSDave Kleikamp if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { 5220b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "External journal has more than one " 5221b31e1552SEric Sandeen "user (unsupported) - %d", 5222ac27a0ecSDave Kleikamp be32_to_cpu(journal->j_superblock->s_nr_users)); 5223ac27a0ecSDave Kleikamp goto out_journal; 5224ac27a0ecSDave Kleikamp } 5225ee7ed3aaSChunguang Xu EXT4_SB(sb)->s_journal_bdev = bdev; 5226617ba13bSMingming Cao ext4_init_journal_params(sb, journal); 5227ac27a0ecSDave Kleikamp return journal; 52280b8e58a1SAndreas Dilger 5229ac27a0ecSDave Kleikamp out_journal: 5230dab291afSMingming Cao jbd2_journal_destroy(journal); 5231ac27a0ecSDave Kleikamp out_bdev: 5232617ba13bSMingming Cao ext4_blkdev_put(bdev); 5233ac27a0ecSDave Kleikamp return NULL; 5234ac27a0ecSDave Kleikamp } 5235ac27a0ecSDave Kleikamp 5236617ba13bSMingming Cao static int ext4_load_journal(struct super_block *sb, 5237617ba13bSMingming Cao struct ext4_super_block *es, 5238ac27a0ecSDave Kleikamp unsigned long journal_devnum) 5239ac27a0ecSDave Kleikamp { 5240ac27a0ecSDave Kleikamp journal_t *journal; 5241ac27a0ecSDave Kleikamp unsigned int journal_inum = le32_to_cpu(es->s_journal_inum); 5242ac27a0ecSDave Kleikamp dev_t journal_dev; 5243ac27a0ecSDave Kleikamp int err = 0; 5244ac27a0ecSDave Kleikamp int really_read_only; 5245273108faSLukas Czerner int journal_dev_ro; 5246ac27a0ecSDave Kleikamp 524711215630SJan Kara if (WARN_ON_ONCE(!ext4_has_feature_journal(sb))) 524811215630SJan Kara return -EFSCORRUPTED; 52490390131bSFrank Mayhar 5250ac27a0ecSDave Kleikamp if (journal_devnum && 5251ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5252b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "external journal device major/minor " 5253b31e1552SEric Sandeen "numbers have changed"); 5254ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(journal_devnum); 5255ac27a0ecSDave Kleikamp } else 5256ac27a0ecSDave Kleikamp journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev)); 5257ac27a0ecSDave Kleikamp 5258273108faSLukas Czerner if (journal_inum && journal_dev) { 5259273108faSLukas Czerner ext4_msg(sb, KERN_ERR, 5260273108faSLukas Czerner "filesystem has both journal inode and journal device!"); 5261273108faSLukas Czerner return -EINVAL; 5262273108faSLukas Czerner } 5263273108faSLukas Czerner 5264273108faSLukas Czerner if (journal_inum) { 5265273108faSLukas Czerner journal = ext4_get_journal(sb, journal_inum); 5266273108faSLukas Czerner if (!journal) 5267273108faSLukas Czerner return -EINVAL; 5268273108faSLukas Czerner } else { 5269273108faSLukas Czerner journal = ext4_get_dev_journal(sb, journal_dev); 5270273108faSLukas Czerner if (!journal) 5271273108faSLukas Czerner return -EINVAL; 5272273108faSLukas Czerner } 5273273108faSLukas Czerner 5274273108faSLukas Czerner journal_dev_ro = bdev_read_only(journal->j_dev); 5275273108faSLukas Czerner really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro; 5276273108faSLukas Czerner 5277273108faSLukas Czerner if (journal_dev_ro && !sb_rdonly(sb)) { 5278273108faSLukas Czerner ext4_msg(sb, KERN_ERR, 5279273108faSLukas Czerner "journal device read-only, try mounting with '-o ro'"); 5280273108faSLukas Czerner err = -EROFS; 5281273108faSLukas Czerner goto err_out; 5282273108faSLukas Czerner } 5283ac27a0ecSDave Kleikamp 5284ac27a0ecSDave Kleikamp /* 5285ac27a0ecSDave Kleikamp * Are we loading a blank journal or performing recovery after a 5286ac27a0ecSDave Kleikamp * crash? For recovery, we need to check in advance whether we 5287ac27a0ecSDave Kleikamp * can get read-write access to the device. 5288ac27a0ecSDave Kleikamp */ 5289e2b911c5SDarrick J. Wong if (ext4_has_feature_journal_needs_recovery(sb)) { 5290bc98a42cSDavid Howells if (sb_rdonly(sb)) { 5291b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "INFO: recovery " 5292b31e1552SEric Sandeen "required on readonly filesystem"); 5293ac27a0ecSDave Kleikamp if (really_read_only) { 5294b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "write access " 5295d98bf8cdSSimon Ruderich "unavailable, cannot proceed " 5296d98bf8cdSSimon Ruderich "(try mounting with noload)"); 5297273108faSLukas Czerner err = -EROFS; 5298273108faSLukas Czerner goto err_out; 5299ac27a0ecSDave Kleikamp } 5300b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "write access will " 5301b31e1552SEric Sandeen "be enabled during recovery"); 5302ac27a0ecSDave Kleikamp } 5303ac27a0ecSDave Kleikamp } 5304ac27a0ecSDave Kleikamp 530590576c0bSTheodore Ts'o if (!(journal->j_flags & JBD2_BARRIER)) 5306b31e1552SEric Sandeen ext4_msg(sb, KERN_INFO, "barriers disabled"); 53074776004fSTheodore Ts'o 5308e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal_needs_recovery(sb)) 5309dab291afSMingming Cao err = jbd2_journal_wipe(journal, !really_read_only); 53101c13d5c0STheodore Ts'o if (!err) { 53111c13d5c0STheodore Ts'o char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL); 53121c13d5c0STheodore Ts'o if (save) 53131c13d5c0STheodore Ts'o memcpy(save, ((char *) es) + 53141c13d5c0STheodore Ts'o EXT4_S_ERR_START, EXT4_S_ERR_LEN); 5315dab291afSMingming Cao err = jbd2_journal_load(journal); 53161c13d5c0STheodore Ts'o if (save) 53171c13d5c0STheodore Ts'o memcpy(((char *) es) + EXT4_S_ERR_START, 53181c13d5c0STheodore Ts'o save, EXT4_S_ERR_LEN); 53191c13d5c0STheodore Ts'o kfree(save); 53201c13d5c0STheodore Ts'o } 5321ac27a0ecSDave Kleikamp 5322ac27a0ecSDave Kleikamp if (err) { 5323b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "error loading journal"); 5324273108faSLukas Czerner goto err_out; 5325ac27a0ecSDave Kleikamp } 5326ac27a0ecSDave Kleikamp 5327617ba13bSMingming Cao EXT4_SB(sb)->s_journal = journal; 532811215630SJan Kara err = ext4_clear_journal_err(sb, es); 532911215630SJan Kara if (err) { 533011215630SJan Kara EXT4_SB(sb)->s_journal = NULL; 533111215630SJan Kara jbd2_journal_destroy(journal); 533211215630SJan Kara return err; 533311215630SJan Kara } 5334ac27a0ecSDave Kleikamp 5335c41303ceSMaciej Żenczykowski if (!really_read_only && journal_devnum && 5336ac27a0ecSDave Kleikamp journal_devnum != le32_to_cpu(es->s_journal_dev)) { 5337ac27a0ecSDave Kleikamp es->s_journal_dev = cpu_to_le32(journal_devnum); 5338ac27a0ecSDave Kleikamp 5339ac27a0ecSDave Kleikamp /* Make sure we flush the recovery flag to disk. */ 5340e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5341ac27a0ecSDave Kleikamp } 5342ac27a0ecSDave Kleikamp 5343ac27a0ecSDave Kleikamp return 0; 5344273108faSLukas Czerner 5345273108faSLukas Czerner err_out: 5346273108faSLukas Czerner jbd2_journal_destroy(journal); 5347273108faSLukas Czerner return err; 5348ac27a0ecSDave Kleikamp } 5349ac27a0ecSDave Kleikamp 5350e2d67052STheodore Ts'o static int ext4_commit_super(struct super_block *sb, int sync) 5351ac27a0ecSDave Kleikamp { 5352e2d67052STheodore Ts'o struct ext4_super_block *es = EXT4_SB(sb)->s_es; 5353617ba13bSMingming Cao struct buffer_head *sbh = EXT4_SB(sb)->s_sbh; 5354c4be0c1dSTakashi Sato int error = 0; 5355ac27a0ecSDave Kleikamp 5356bdfe0cbdSTheodore Ts'o if (!sbh || block_device_ejected(sb)) 5357c4be0c1dSTakashi Sato return error; 5358a17712c8SJon Derrick 5359a17712c8SJon Derrick /* 536071290b36STheodore Ts'o * If the file system is mounted read-only, don't update the 536171290b36STheodore Ts'o * superblock write time. This avoids updating the superblock 536271290b36STheodore Ts'o * write time when we are mounting the root file system 536371290b36STheodore Ts'o * read/only but we need to replay the journal; at that point, 536471290b36STheodore Ts'o * for people who are east of GMT and who make their clock 536571290b36STheodore Ts'o * tick in localtime for Windows bug-for-bug compatibility, 536671290b36STheodore Ts'o * the clock is set in the future, and this will cause e2fsck 536771290b36STheodore Ts'o * to complain and force a full file system check. 536871290b36STheodore Ts'o */ 53691751e8a6SLinus Torvalds if (!(sb->s_flags & SB_RDONLY)) 53706a0678a7SArnd Bergmann ext4_update_tstamp(es, s_wtime); 5371f613dfcbSTheodore Ts'o if (sb->s_bdev->bd_part) 5372afc32f7eSTheodore Ts'o es->s_kbytes_written = 5373afc32f7eSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 5374dbae2c55SMichael Callahan ((part_stat_read(sb->s_bdev->bd_part, 5375dbae2c55SMichael Callahan sectors[STAT_WRITE]) - 5376afc32f7eSTheodore Ts'o EXT4_SB(sb)->s_sectors_written_start) >> 1)); 5377f613dfcbSTheodore Ts'o else 5378f613dfcbSTheodore Ts'o es->s_kbytes_written = 5379f613dfcbSTheodore Ts'o cpu_to_le64(EXT4_SB(sb)->s_kbytes_written); 5380d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter)) 538157042651STheodore Ts'o ext4_free_blocks_count_set(es, 538257042651STheodore Ts'o EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive( 538357042651STheodore Ts'o &EXT4_SB(sb)->s_freeclusters_counter))); 5384d5e03cbbSTheodore Ts'o if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter)) 53857f93cff9STheodore Ts'o es->s_free_inodes_count = 53867f93cff9STheodore Ts'o cpu_to_le32(percpu_counter_sum_positive( 53875d1b1b3fSAneesh Kumar K.V &EXT4_SB(sb)->s_freeinodes_counter)); 5388ac27a0ecSDave Kleikamp BUFFER_TRACE(sbh, "marking dirty"); 538906db49e6STheodore Ts'o ext4_superblock_csum_set(sb); 53901566a48aSTheodore Ts'o if (sync) 53914743f839SPranay Kr. Srivastava lock_buffer(sbh); 5392e8680786STheodore Ts'o if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) { 53934743f839SPranay Kr. Srivastava /* 53944743f839SPranay Kr. Srivastava * Oh, dear. A previous attempt to write the 53954743f839SPranay Kr. Srivastava * superblock failed. This could happen because the 53964743f839SPranay Kr. Srivastava * USB device was yanked out. Or it could happen to 53974743f839SPranay Kr. Srivastava * be a transient write error and maybe the block will 53984743f839SPranay Kr. Srivastava * be remapped. Nothing we can do but to retry the 53994743f839SPranay Kr. Srivastava * write and hope for the best. 54004743f839SPranay Kr. Srivastava */ 54014743f839SPranay Kr. Srivastava ext4_msg(sb, KERN_ERR, "previous I/O error to " 54024743f839SPranay Kr. Srivastava "superblock detected"); 54034743f839SPranay Kr. Srivastava clear_buffer_write_io_error(sbh); 54044743f839SPranay Kr. Srivastava set_buffer_uptodate(sbh); 54054743f839SPranay Kr. Srivastava } 5406ac27a0ecSDave Kleikamp mark_buffer_dirty(sbh); 5407914258bfSTheodore Ts'o if (sync) { 54081566a48aSTheodore Ts'o unlock_buffer(sbh); 5409564bc402SDaeho Jeong error = __sync_dirty_buffer(sbh, 541000473374SJan Kara REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0)); 5411c89128a0SJaegeuk Kim if (buffer_write_io_error(sbh)) { 5412b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, "I/O error while writing " 5413b31e1552SEric Sandeen "superblock"); 5414914258bfSTheodore Ts'o clear_buffer_write_io_error(sbh); 5415914258bfSTheodore Ts'o set_buffer_uptodate(sbh); 5416914258bfSTheodore Ts'o } 5417914258bfSTheodore Ts'o } 5418c4be0c1dSTakashi Sato return error; 5419ac27a0ecSDave Kleikamp } 5420ac27a0ecSDave Kleikamp 5421ac27a0ecSDave Kleikamp /* 5422ac27a0ecSDave Kleikamp * Have we just finished recovery? If so, and if we are mounting (or 5423ac27a0ecSDave Kleikamp * remounting) the filesystem readonly, then we will end up with a 5424ac27a0ecSDave Kleikamp * consistent fs on disk. Record that fact. 5425ac27a0ecSDave Kleikamp */ 542611215630SJan Kara static int ext4_mark_recovery_complete(struct super_block *sb, 5427617ba13bSMingming Cao struct ext4_super_block *es) 5428ac27a0ecSDave Kleikamp { 542911215630SJan Kara int err; 5430617ba13bSMingming Cao journal_t *journal = EXT4_SB(sb)->s_journal; 5431ac27a0ecSDave Kleikamp 5432e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) { 543311215630SJan Kara if (journal != NULL) { 543411215630SJan Kara ext4_error(sb, "Journal got removed while the fs was " 543511215630SJan Kara "mounted!"); 543611215630SJan Kara return -EFSCORRUPTED; 543711215630SJan Kara } 543811215630SJan Kara return 0; 54390390131bSFrank Mayhar } 5440dab291afSMingming Cao jbd2_journal_lock_updates(journal); 544111215630SJan Kara err = jbd2_journal_flush(journal); 544211215630SJan Kara if (err < 0) 54437ffe1ea8SHidehiro Kawai goto out; 54447ffe1ea8SHidehiro Kawai 5445bc98a42cSDavid Howells if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) { 5446e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5447e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5448ac27a0ecSDave Kleikamp } 54497ffe1ea8SHidehiro Kawai out: 5450dab291afSMingming Cao jbd2_journal_unlock_updates(journal); 545111215630SJan Kara return err; 5452ac27a0ecSDave Kleikamp } 5453ac27a0ecSDave Kleikamp 5454ac27a0ecSDave Kleikamp /* 5455ac27a0ecSDave Kleikamp * If we are mounting (or read-write remounting) a filesystem whose journal 5456ac27a0ecSDave Kleikamp * has recorded an error from a previous lifetime, move that error to the 5457ac27a0ecSDave Kleikamp * main filesystem now. 5458ac27a0ecSDave Kleikamp */ 545911215630SJan Kara static int ext4_clear_journal_err(struct super_block *sb, 5460617ba13bSMingming Cao struct ext4_super_block *es) 5461ac27a0ecSDave Kleikamp { 5462ac27a0ecSDave Kleikamp journal_t *journal; 5463ac27a0ecSDave Kleikamp int j_errno; 5464ac27a0ecSDave Kleikamp const char *errstr; 5465ac27a0ecSDave Kleikamp 546611215630SJan Kara if (!ext4_has_feature_journal(sb)) { 546711215630SJan Kara ext4_error(sb, "Journal got removed while the fs was mounted!"); 546811215630SJan Kara return -EFSCORRUPTED; 546911215630SJan Kara } 54700390131bSFrank Mayhar 5471617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5472ac27a0ecSDave Kleikamp 5473ac27a0ecSDave Kleikamp /* 5474ac27a0ecSDave Kleikamp * Now check for any error status which may have been recorded in the 5475617ba13bSMingming Cao * journal by a prior ext4_error() or ext4_abort() 5476ac27a0ecSDave Kleikamp */ 5477ac27a0ecSDave Kleikamp 5478dab291afSMingming Cao j_errno = jbd2_journal_errno(journal); 5479ac27a0ecSDave Kleikamp if (j_errno) { 5480ac27a0ecSDave Kleikamp char nbuf[16]; 5481ac27a0ecSDave Kleikamp 5482617ba13bSMingming Cao errstr = ext4_decode_error(sb, j_errno, nbuf); 548312062dddSEric Sandeen ext4_warning(sb, "Filesystem error recorded " 5484ac27a0ecSDave Kleikamp "from previous mount: %s", errstr); 548512062dddSEric Sandeen ext4_warning(sb, "Marking fs in need of filesystem check."); 5486ac27a0ecSDave Kleikamp 5487617ba13bSMingming Cao EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 5488617ba13bSMingming Cao es->s_state |= cpu_to_le16(EXT4_ERROR_FS); 5489e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5490ac27a0ecSDave Kleikamp 5491dab291afSMingming Cao jbd2_journal_clear_err(journal); 5492d796c52eSTheodore Ts'o jbd2_journal_update_sb_errno(journal); 5493ac27a0ecSDave Kleikamp } 549411215630SJan Kara return 0; 5495ac27a0ecSDave Kleikamp } 5496ac27a0ecSDave Kleikamp 5497ac27a0ecSDave Kleikamp /* 5498ac27a0ecSDave Kleikamp * Force the running and committing transactions to commit, 5499ac27a0ecSDave Kleikamp * and wait on the commit. 5500ac27a0ecSDave Kleikamp */ 5501617ba13bSMingming Cao int ext4_force_commit(struct super_block *sb) 5502ac27a0ecSDave Kleikamp { 5503ac27a0ecSDave Kleikamp journal_t *journal; 5504ac27a0ecSDave Kleikamp 5505bc98a42cSDavid Howells if (sb_rdonly(sb)) 5506ac27a0ecSDave Kleikamp return 0; 5507ac27a0ecSDave Kleikamp 5508617ba13bSMingming Cao journal = EXT4_SB(sb)->s_journal; 5509b1deefc9SGuo Chao return ext4_journal_force_commit(journal); 5510ac27a0ecSDave Kleikamp } 5511ac27a0ecSDave Kleikamp 5512617ba13bSMingming Cao static int ext4_sync_fs(struct super_block *sb, int wait) 5513ac27a0ecSDave Kleikamp { 551414ce0cb4STheodore Ts'o int ret = 0; 55159eddacf9SJan Kara tid_t target; 551606a407f1SDmitry Monakhov bool needs_barrier = false; 55178d5d02e6SMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5518ac27a0ecSDave Kleikamp 551949598e04SJun Piao if (unlikely(ext4_forced_shutdown(sbi))) 55200db1ff22STheodore Ts'o return 0; 55210db1ff22STheodore Ts'o 55229bffad1eSTheodore Ts'o trace_ext4_sync_fs(sb, wait); 55232e8fa54eSJan Kara flush_workqueue(sbi->rsv_conversion_wq); 5524a1177825SJan Kara /* 5525a1177825SJan Kara * Writeback quota in non-journalled quota case - journalled quota has 5526a1177825SJan Kara * no dirty dquots 5527a1177825SJan Kara */ 5528a1177825SJan Kara dquot_writeback_dquots(sb, -1); 552906a407f1SDmitry Monakhov /* 553006a407f1SDmitry Monakhov * Data writeback is possible w/o journal transaction, so barrier must 553106a407f1SDmitry Monakhov * being sent at the end of the function. But we can skip it if 553206a407f1SDmitry Monakhov * transaction_commit will do it for us. 553306a407f1SDmitry Monakhov */ 5534bda32530STheodore Ts'o if (sbi->s_journal) { 553506a407f1SDmitry Monakhov target = jbd2_get_latest_transaction(sbi->s_journal); 553606a407f1SDmitry Monakhov if (wait && sbi->s_journal->j_flags & JBD2_BARRIER && 553706a407f1SDmitry Monakhov !jbd2_trans_will_send_data_barrier(sbi->s_journal, target)) 553806a407f1SDmitry Monakhov needs_barrier = true; 553906a407f1SDmitry Monakhov 55408d5d02e6SMingming Cao if (jbd2_journal_start_commit(sbi->s_journal, &target)) { 5541ac27a0ecSDave Kleikamp if (wait) 5542bda32530STheodore Ts'o ret = jbd2_log_wait_commit(sbi->s_journal, 5543bda32530STheodore Ts'o target); 55440390131bSFrank Mayhar } 5545bda32530STheodore Ts'o } else if (wait && test_opt(sb, BARRIER)) 5546bda32530STheodore Ts'o needs_barrier = true; 554706a407f1SDmitry Monakhov if (needs_barrier) { 554806a407f1SDmitry Monakhov int err; 55499398554fSChristoph Hellwig err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL); 555006a407f1SDmitry Monakhov if (!ret) 555106a407f1SDmitry Monakhov ret = err; 555206a407f1SDmitry Monakhov } 555306a407f1SDmitry Monakhov 555406a407f1SDmitry Monakhov return ret; 555506a407f1SDmitry Monakhov } 555606a407f1SDmitry Monakhov 5557ac27a0ecSDave Kleikamp /* 5558ac27a0ecSDave Kleikamp * LVM calls this function before a (read-only) snapshot is created. This 5559ac27a0ecSDave Kleikamp * gives us a chance to flush the journal completely and mark the fs clean. 5560be4f27d3SYongqiang Yang * 5561be4f27d3SYongqiang Yang * Note that only this function cannot bring a filesystem to be in a clean 55628e8ad8a5SJan Kara * state independently. It relies on upper layer to stop all data & metadata 55638e8ad8a5SJan Kara * modifications. 5564ac27a0ecSDave Kleikamp */ 5565c4be0c1dSTakashi Sato static int ext4_freeze(struct super_block *sb) 5566ac27a0ecSDave Kleikamp { 5567c4be0c1dSTakashi Sato int error = 0; 5568c4be0c1dSTakashi Sato journal_t *journal; 5569ac27a0ecSDave Kleikamp 5570bc98a42cSDavid Howells if (sb_rdonly(sb)) 55719ca92389STheodore Ts'o return 0; 55729ca92389STheodore Ts'o 5573c4be0c1dSTakashi Sato journal = EXT4_SB(sb)->s_journal; 5574ac27a0ecSDave Kleikamp 5575bb044576STheodore Ts'o if (journal) { 5576ac27a0ecSDave Kleikamp /* Now we set up the journal barrier. */ 5577dab291afSMingming Cao jbd2_journal_lock_updates(journal); 55787ffe1ea8SHidehiro Kawai 55797ffe1ea8SHidehiro Kawai /* 5580bb044576STheodore Ts'o * Don't clear the needs_recovery flag if we failed to 5581bb044576STheodore Ts'o * flush the journal. 55827ffe1ea8SHidehiro Kawai */ 5583c4be0c1dSTakashi Sato error = jbd2_journal_flush(journal); 55846b0310fbSEric Sandeen if (error < 0) 55856b0310fbSEric Sandeen goto out; 5586ac27a0ecSDave Kleikamp 5587ac27a0ecSDave Kleikamp /* Journal blocked and flushed, clear needs_recovery flag. */ 5588e2b911c5SDarrick J. Wong ext4_clear_feature_journal_needs_recovery(sb); 5589c642dc9eSEric Sandeen } 5590c642dc9eSEric Sandeen 5591e2d67052STheodore Ts'o error = ext4_commit_super(sb, 1); 55926b0310fbSEric Sandeen out: 5593bb044576STheodore Ts'o if (journal) 55948e8ad8a5SJan Kara /* we rely on upper layer to stop further updates */ 5595bb044576STheodore Ts'o jbd2_journal_unlock_updates(journal); 55966b0310fbSEric Sandeen return error; 5597ac27a0ecSDave Kleikamp } 5598ac27a0ecSDave Kleikamp 5599ac27a0ecSDave Kleikamp /* 5600ac27a0ecSDave Kleikamp * Called by LVM after the snapshot is done. We need to reset the RECOVER 5601ac27a0ecSDave Kleikamp * flag here, even though the filesystem is not technically dirty yet. 5602ac27a0ecSDave Kleikamp */ 5603c4be0c1dSTakashi Sato static int ext4_unfreeze(struct super_block *sb) 5604ac27a0ecSDave Kleikamp { 5605bc98a42cSDavid Howells if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb))) 56069ca92389STheodore Ts'o return 0; 56079ca92389STheodore Ts'o 5608c642dc9eSEric Sandeen if (EXT4_SB(sb)->s_journal) { 56099ca92389STheodore Ts'o /* Reset the needs_recovery flag before the fs is unlocked. */ 5610e2b911c5SDarrick J. Wong ext4_set_feature_journal_needs_recovery(sb); 5611c642dc9eSEric Sandeen } 5612c642dc9eSEric Sandeen 5613e2d67052STheodore Ts'o ext4_commit_super(sb, 1); 5614c4be0c1dSTakashi Sato return 0; 5615ac27a0ecSDave Kleikamp } 5616ac27a0ecSDave Kleikamp 5617673c6100STheodore Ts'o /* 5618673c6100STheodore Ts'o * Structure to save mount options for ext4_remount's benefit 5619673c6100STheodore Ts'o */ 5620673c6100STheodore Ts'o struct ext4_mount_options { 5621673c6100STheodore Ts'o unsigned long s_mount_opt; 5622a2595b8aSTheodore Ts'o unsigned long s_mount_opt2; 562308cefc7aSEric W. Biederman kuid_t s_resuid; 562408cefc7aSEric W. Biederman kgid_t s_resgid; 5625673c6100STheodore Ts'o unsigned long s_commit_interval; 5626673c6100STheodore Ts'o u32 s_min_batch_time, s_max_batch_time; 5627673c6100STheodore Ts'o #ifdef CONFIG_QUOTA 5628673c6100STheodore Ts'o int s_jquota_fmt; 5629a2d4a646SJan Kara char *s_qf_names[EXT4_MAXQUOTAS]; 5630673c6100STheodore Ts'o #endif 5631673c6100STheodore Ts'o }; 5632673c6100STheodore Ts'o 5633617ba13bSMingming Cao static int ext4_remount(struct super_block *sb, int *flags, char *data) 5634ac27a0ecSDave Kleikamp { 5635617ba13bSMingming Cao struct ext4_super_block *es; 5636617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5637f25391ebSLukas Czerner unsigned long old_sb_flags, vfs_flags; 5638617ba13bSMingming Cao struct ext4_mount_options old_opts; 5639c79d967dSChristoph Hellwig int enable_quota = 0; 56408a266467STheodore Ts'o ext4_group_t g; 5641b3881f74STheodore Ts'o unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 5642c5e06d10SJohann Lombardi int err = 0; 5643ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 564403dafb5fSChen Gang int i, j; 564533458eabSTheodore Ts'o char *to_free[EXT4_MAXQUOTAS]; 5646ac27a0ecSDave Kleikamp #endif 5647d4c402d9SCurt Wohlgemuth char *orig_data = kstrdup(data, GFP_KERNEL); 5648ac27a0ecSDave Kleikamp 564921ac738eSChengguang Xu if (data && !orig_data) 565021ac738eSChengguang Xu return -ENOMEM; 565121ac738eSChengguang Xu 5652ac27a0ecSDave Kleikamp /* Store the original options */ 5653ac27a0ecSDave Kleikamp old_sb_flags = sb->s_flags; 5654ac27a0ecSDave Kleikamp old_opts.s_mount_opt = sbi->s_mount_opt; 5655a2595b8aSTheodore Ts'o old_opts.s_mount_opt2 = sbi->s_mount_opt2; 5656ac27a0ecSDave Kleikamp old_opts.s_resuid = sbi->s_resuid; 5657ac27a0ecSDave Kleikamp old_opts.s_resgid = sbi->s_resgid; 5658ac27a0ecSDave Kleikamp old_opts.s_commit_interval = sbi->s_commit_interval; 565930773840STheodore Ts'o old_opts.s_min_batch_time = sbi->s_min_batch_time; 566030773840STheodore Ts'o old_opts.s_max_batch_time = sbi->s_max_batch_time; 5661ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5662ac27a0ecSDave Kleikamp old_opts.s_jquota_fmt = sbi->s_jquota_fmt; 5663a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 566403dafb5fSChen Gang if (sbi->s_qf_names[i]) { 566533458eabSTheodore Ts'o char *qf_name = get_qf_name(sb, sbi, i); 566633458eabSTheodore Ts'o 566733458eabSTheodore Ts'o old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL); 566803dafb5fSChen Gang if (!old_opts.s_qf_names[i]) { 566903dafb5fSChen Gang for (j = 0; j < i; j++) 567003dafb5fSChen Gang kfree(old_opts.s_qf_names[j]); 56713e36a163SWei Yongjun kfree(orig_data); 567203dafb5fSChen Gang return -ENOMEM; 567303dafb5fSChen Gang } 567403dafb5fSChen Gang } else 567503dafb5fSChen Gang old_opts.s_qf_names[i] = NULL; 5676ac27a0ecSDave Kleikamp #endif 5677b3881f74STheodore Ts'o if (sbi->s_journal && sbi->s_journal->j_task->io_context) 5678b3881f74STheodore Ts'o journal_ioprio = sbi->s_journal->j_task->io_context->ioprio; 5679ac27a0ecSDave Kleikamp 5680f25391ebSLukas Czerner /* 5681f25391ebSLukas Czerner * Some options can be enabled by ext4 and/or by VFS mount flag 5682f25391ebSLukas Czerner * either way we need to make sure it matches in both *flags and 5683f25391ebSLukas Czerner * s_flags. Copy those selected flags from *flags to s_flags 5684f25391ebSLukas Czerner */ 5685f25391ebSLukas Czerner vfs_flags = SB_LAZYTIME | SB_I_VERSION; 5686f25391ebSLukas Czerner sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags); 5687f25391ebSLukas Czerner 5688661aa520SEric Sandeen if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) { 5689ac27a0ecSDave Kleikamp err = -EINVAL; 5690ac27a0ecSDave Kleikamp goto restore_opts; 5691ac27a0ecSDave Kleikamp } 5692ac27a0ecSDave Kleikamp 56936b992ff2SDarrick J. Wong if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^ 56946b992ff2SDarrick J. Wong test_opt(sb, JOURNAL_CHECKSUM)) { 56956b992ff2SDarrick J. Wong ext4_msg(sb, KERN_ERR, "changing journal_checksum " 56962d5b86e0SEric Sandeen "during remount not supported; ignoring"); 56972d5b86e0SEric Sandeen sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM; 5698c6d3d56dSDarrick J. Wong } 5699c6d3d56dSDarrick J. Wong 57006ae6514bSPiotr Sarna if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) { 57016ae6514bSPiotr Sarna if (test_opt2(sb, EXPLICIT_DELALLOC)) { 57026ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 57036ae6514bSPiotr Sarna "both data=journal and delalloc"); 57046ae6514bSPiotr Sarna err = -EINVAL; 57056ae6514bSPiotr Sarna goto restore_opts; 57066ae6514bSPiotr Sarna } 57076ae6514bSPiotr Sarna if (test_opt(sb, DIOREAD_NOLOCK)) { 57086ae6514bSPiotr Sarna ext4_msg(sb, KERN_ERR, "can't mount with " 57096ae6514bSPiotr Sarna "both data=journal and dioread_nolock"); 57106ae6514bSPiotr Sarna err = -EINVAL; 57116ae6514bSPiotr Sarna goto restore_opts; 57126ae6514bSPiotr Sarna } 5713ab04df78SJan Kara } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) { 5714ab04df78SJan Kara if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) { 5715ab04df78SJan Kara ext4_msg(sb, KERN_ERR, "can't mount with " 5716ab04df78SJan Kara "journal_async_commit in data=ordered mode"); 5717ab04df78SJan Kara err = -EINVAL; 5718ab04df78SJan Kara goto restore_opts; 5719ab04df78SJan Kara } 5720923ae0ffSRoss Zwisler } 5721923ae0ffSRoss Zwisler 5722cdb7ee4cSTahsin Erdogan if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) { 5723cdb7ee4cSTahsin Erdogan ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount"); 5724cdb7ee4cSTahsin Erdogan err = -EINVAL; 5725cdb7ee4cSTahsin Erdogan goto restore_opts; 5726cdb7ee4cSTahsin Erdogan } 5727cdb7ee4cSTahsin Erdogan 57284ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 572954d3adbcSTheodore Ts'o ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user"); 5730ac27a0ecSDave Kleikamp 57311751e8a6SLinus Torvalds sb->s_flags = (sb->s_flags & ~SB_POSIXACL) | 57321751e8a6SLinus Torvalds (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0); 5733ac27a0ecSDave Kleikamp 5734ac27a0ecSDave Kleikamp es = sbi->s_es; 5735ac27a0ecSDave Kleikamp 5736b3881f74STheodore Ts'o if (sbi->s_journal) { 5737617ba13bSMingming Cao ext4_init_journal_params(sb, sbi->s_journal); 5738b3881f74STheodore Ts'o set_task_ioprio(sbi->s_journal->j_task, journal_ioprio); 5739b3881f74STheodore Ts'o } 5740ac27a0ecSDave Kleikamp 57411751e8a6SLinus Torvalds if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) { 57424ab2f15bSTheodore Ts'o if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) { 5743ac27a0ecSDave Kleikamp err = -EROFS; 5744ac27a0ecSDave Kleikamp goto restore_opts; 5745ac27a0ecSDave Kleikamp } 5746ac27a0ecSDave Kleikamp 57471751e8a6SLinus Torvalds if (*flags & SB_RDONLY) { 574838c03b34STheodore Ts'o err = sync_filesystem(sb); 574938c03b34STheodore Ts'o if (err < 0) 575038c03b34STheodore Ts'o goto restore_opts; 57510f0dd62fSChristoph Hellwig err = dquot_suspend(sb, -1); 57520f0dd62fSChristoph Hellwig if (err < 0) 5753c79d967dSChristoph Hellwig goto restore_opts; 5754c79d967dSChristoph Hellwig 5755ac27a0ecSDave Kleikamp /* 5756ac27a0ecSDave Kleikamp * First of all, the unconditional stuff we have to do 5757ac27a0ecSDave Kleikamp * to disable replay of the journal when we next remount 5758ac27a0ecSDave Kleikamp */ 57591751e8a6SLinus Torvalds sb->s_flags |= SB_RDONLY; 5760ac27a0ecSDave Kleikamp 5761ac27a0ecSDave Kleikamp /* 5762ac27a0ecSDave Kleikamp * OK, test if we are remounting a valid rw partition 5763ac27a0ecSDave Kleikamp * readonly, and if so set the rdonly flag and then 5764ac27a0ecSDave Kleikamp * mark the partition as valid again. 5765ac27a0ecSDave Kleikamp */ 5766617ba13bSMingming Cao if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) && 5767617ba13bSMingming Cao (sbi->s_mount_state & EXT4_VALID_FS)) 5768ac27a0ecSDave Kleikamp es->s_state = cpu_to_le16(sbi->s_mount_state); 5769ac27a0ecSDave Kleikamp 577011215630SJan Kara if (sbi->s_journal) { 577111215630SJan Kara /* 577211215630SJan Kara * We let remount-ro finish even if marking fs 577311215630SJan Kara * as clean failed... 577411215630SJan Kara */ 5775617ba13bSMingming Cao ext4_mark_recovery_complete(sb, es); 577611215630SJan Kara } 57772dca60d9STheodore Ts'o if (sbi->s_mmp_tsk) 57782dca60d9STheodore Ts'o kthread_stop(sbi->s_mmp_tsk); 5779ac27a0ecSDave Kleikamp } else { 5780a13fb1a4SEric Sandeen /* Make sure we can mount this feature set readwrite */ 5781e2b911c5SDarrick J. Wong if (ext4_has_feature_readonly(sb) || 57822cb5cc8bSDarrick J. Wong !ext4_feature_set_ok(sb, 0)) { 5783ac27a0ecSDave Kleikamp err = -EROFS; 5784ac27a0ecSDave Kleikamp goto restore_opts; 5785ac27a0ecSDave Kleikamp } 5786ead6596bSEric Sandeen /* 57878a266467STheodore Ts'o * Make sure the group descriptor checksums 57880b8e58a1SAndreas Dilger * are sane. If they aren't, refuse to remount r/w. 57898a266467STheodore Ts'o */ 57908a266467STheodore Ts'o for (g = 0; g < sbi->s_groups_count; g++) { 57918a266467STheodore Ts'o struct ext4_group_desc *gdp = 57928a266467STheodore Ts'o ext4_get_group_desc(sb, g, NULL); 57938a266467STheodore Ts'o 5794feb0ab32SDarrick J. Wong if (!ext4_group_desc_csum_verify(sb, g, gdp)) { 5795b31e1552SEric Sandeen ext4_msg(sb, KERN_ERR, 5796b31e1552SEric Sandeen "ext4_remount: Checksum for group %u failed (%u!=%u)", 5797e2b911c5SDarrick J. Wong g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)), 57988a266467STheodore Ts'o le16_to_cpu(gdp->bg_checksum)); 57996a797d27SDarrick J. Wong err = -EFSBADCRC; 58008a266467STheodore Ts'o goto restore_opts; 58018a266467STheodore Ts'o } 58028a266467STheodore Ts'o } 58038a266467STheodore Ts'o 58048a266467STheodore Ts'o /* 5805ead6596bSEric Sandeen * If we have an unprocessed orphan list hanging 5806ead6596bSEric Sandeen * around from a previously readonly bdev mount, 5807ead6596bSEric Sandeen * require a full umount/remount for now. 5808ead6596bSEric Sandeen */ 5809ead6596bSEric Sandeen if (es->s_last_orphan) { 5810b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Couldn't " 5811ead6596bSEric Sandeen "remount RDWR because of unprocessed " 5812ead6596bSEric Sandeen "orphan inode list. Please " 5813b31e1552SEric Sandeen "umount/remount instead"); 5814ead6596bSEric Sandeen err = -EINVAL; 5815ead6596bSEric Sandeen goto restore_opts; 5816ead6596bSEric Sandeen } 5817ead6596bSEric Sandeen 5818ac27a0ecSDave Kleikamp /* 5819ac27a0ecSDave Kleikamp * Mounting a RDONLY partition read-write, so reread 5820ac27a0ecSDave Kleikamp * and store the current valid flag. (It may have 5821ac27a0ecSDave Kleikamp * been changed by e2fsck since we originally mounted 5822ac27a0ecSDave Kleikamp * the partition.) 5823ac27a0ecSDave Kleikamp */ 582411215630SJan Kara if (sbi->s_journal) { 582511215630SJan Kara err = ext4_clear_journal_err(sb, es); 582611215630SJan Kara if (err) 582711215630SJan Kara goto restore_opts; 582811215630SJan Kara } 5829ac27a0ecSDave Kleikamp sbi->s_mount_state = le16_to_cpu(es->s_state); 5830c89128a0SJaegeuk Kim 5831c89128a0SJaegeuk Kim err = ext4_setup_super(sb, es, 0); 5832c89128a0SJaegeuk Kim if (err) 5833c89128a0SJaegeuk Kim goto restore_opts; 5834c89128a0SJaegeuk Kim 58351751e8a6SLinus Torvalds sb->s_flags &= ~SB_RDONLY; 5836e2b911c5SDarrick J. Wong if (ext4_has_feature_mmp(sb)) 5837c5e06d10SJohann Lombardi if (ext4_multi_mount_protect(sb, 5838c5e06d10SJohann Lombardi le64_to_cpu(es->s_mmp_block))) { 5839c5e06d10SJohann Lombardi err = -EROFS; 5840c5e06d10SJohann Lombardi goto restore_opts; 5841c5e06d10SJohann Lombardi } 5842c79d967dSChristoph Hellwig enable_quota = 1; 5843ac27a0ecSDave Kleikamp } 5844ac27a0ecSDave Kleikamp } 5845bfff6873SLukas Czerner 5846bfff6873SLukas Czerner /* 5847bfff6873SLukas Czerner * Reinitialize lazy itable initialization thread based on 5848bfff6873SLukas Czerner * current settings 5849bfff6873SLukas Czerner */ 5850bc98a42cSDavid Howells if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE)) 5851bfff6873SLukas Czerner ext4_unregister_li_request(sb); 5852bfff6873SLukas Czerner else { 5853bfff6873SLukas Czerner ext4_group_t first_not_zeroed; 5854bfff6873SLukas Czerner first_not_zeroed = ext4_has_uninit_itable(sb); 5855bfff6873SLukas Czerner ext4_register_li_request(sb, first_not_zeroed); 5856bfff6873SLukas Czerner } 5857bfff6873SLukas Czerner 58580f5bde1dSJan Kara /* 58590f5bde1dSJan Kara * Handle creation of system zone data early because it can fail. 58600f5bde1dSJan Kara * Releasing of existing data is done when we are sure remount will 58610f5bde1dSJan Kara * succeed. 58620f5bde1dSJan Kara */ 5863dd0db94fSChunguang Xu if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) { 5864d176b1f6SJan Kara err = ext4_setup_system_zone(sb); 5865d176b1f6SJan Kara if (err) 5866d176b1f6SJan Kara goto restore_opts; 58670f5bde1dSJan Kara } 5868d176b1f6SJan Kara 5869c89128a0SJaegeuk Kim if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) { 5870c89128a0SJaegeuk Kim err = ext4_commit_super(sb, 1); 5871c89128a0SJaegeuk Kim if (err) 5872c89128a0SJaegeuk Kim goto restore_opts; 5873c89128a0SJaegeuk Kim } 58740390131bSFrank Mayhar 5875ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5876ac27a0ecSDave Kleikamp /* Release old quota file names */ 5877a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) 5878ac27a0ecSDave Kleikamp kfree(old_opts.s_qf_names[i]); 58797c319d32SAditya Kali if (enable_quota) { 58807c319d32SAditya Kali if (sb_any_quota_suspended(sb)) 58810f0dd62fSChristoph Hellwig dquot_resume(sb, -1); 5882e2b911c5SDarrick J. Wong else if (ext4_has_feature_quota(sb)) { 58837c319d32SAditya Kali err = ext4_enable_quotas(sb); 588407724f98STheodore Ts'o if (err) 58857c319d32SAditya Kali goto restore_opts; 58867c319d32SAditya Kali } 58877c319d32SAditya Kali } 58887c319d32SAditya Kali #endif 5889dd0db94fSChunguang Xu if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 58900f5bde1dSJan Kara ext4_release_system_zone(sb); 5891d4c402d9SCurt Wohlgemuth 5892f25391ebSLukas Czerner /* 5893f25391ebSLukas Czerner * Some options can be enabled by ext4 and/or by VFS mount flag 5894f25391ebSLukas Czerner * either way we need to make sure it matches in both *flags and 5895f25391ebSLukas Czerner * s_flags. Copy those selected flags from s_flags to *flags 5896f25391ebSLukas Czerner */ 5897f25391ebSLukas Czerner *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags); 5898f25391ebSLukas Czerner 5899d4c402d9SCurt Wohlgemuth ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data); 5900d4c402d9SCurt Wohlgemuth kfree(orig_data); 5901ac27a0ecSDave Kleikamp return 0; 59020b8e58a1SAndreas Dilger 5903ac27a0ecSDave Kleikamp restore_opts: 5904ac27a0ecSDave Kleikamp sb->s_flags = old_sb_flags; 5905ac27a0ecSDave Kleikamp sbi->s_mount_opt = old_opts.s_mount_opt; 5906a2595b8aSTheodore Ts'o sbi->s_mount_opt2 = old_opts.s_mount_opt2; 5907ac27a0ecSDave Kleikamp sbi->s_resuid = old_opts.s_resuid; 5908ac27a0ecSDave Kleikamp sbi->s_resgid = old_opts.s_resgid; 5909ac27a0ecSDave Kleikamp sbi->s_commit_interval = old_opts.s_commit_interval; 591030773840STheodore Ts'o sbi->s_min_batch_time = old_opts.s_min_batch_time; 591130773840STheodore Ts'o sbi->s_max_batch_time = old_opts.s_max_batch_time; 5912dd0db94fSChunguang Xu if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks) 59130f5bde1dSJan Kara ext4_release_system_zone(sb); 5914ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 5915ac27a0ecSDave Kleikamp sbi->s_jquota_fmt = old_opts.s_jquota_fmt; 5916a2d4a646SJan Kara for (i = 0; i < EXT4_MAXQUOTAS; i++) { 591733458eabSTheodore Ts'o to_free[i] = get_qf_name(sb, sbi, i); 591833458eabSTheodore Ts'o rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]); 5919ac27a0ecSDave Kleikamp } 592033458eabSTheodore Ts'o synchronize_rcu(); 592133458eabSTheodore Ts'o for (i = 0; i < EXT4_MAXQUOTAS; i++) 592233458eabSTheodore Ts'o kfree(to_free[i]); 5923ac27a0ecSDave Kleikamp #endif 5924d4c402d9SCurt Wohlgemuth kfree(orig_data); 5925ac27a0ecSDave Kleikamp return err; 5926ac27a0ecSDave Kleikamp } 5927ac27a0ecSDave Kleikamp 5928689c958cSLi Xi #ifdef CONFIG_QUOTA 5929689c958cSLi Xi static int ext4_statfs_project(struct super_block *sb, 5930689c958cSLi Xi kprojid_t projid, struct kstatfs *buf) 5931689c958cSLi Xi { 5932689c958cSLi Xi struct kqid qid; 5933689c958cSLi Xi struct dquot *dquot; 5934689c958cSLi Xi u64 limit; 5935689c958cSLi Xi u64 curblock; 5936689c958cSLi Xi 5937689c958cSLi Xi qid = make_kqid_projid(projid); 5938689c958cSLi Xi dquot = dqget(sb, qid); 5939689c958cSLi Xi if (IS_ERR(dquot)) 5940689c958cSLi Xi return PTR_ERR(dquot); 59417b9ca4c6SJan Kara spin_lock(&dquot->dq_dqb_lock); 5942689c958cSLi Xi 5943a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit, 5944a08fe66eSChengguang Xu dquot->dq_dqb.dqb_bhardlimit); 594557c32ea4SChengguang Xu limit >>= sb->s_blocksize_bits; 594657c32ea4SChengguang Xu 5947689c958cSLi Xi if (limit && buf->f_blocks > limit) { 5948f06925c7SKonstantin Khlebnikov curblock = (dquot->dq_dqb.dqb_curspace + 5949f06925c7SKonstantin Khlebnikov dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits; 5950689c958cSLi Xi buf->f_blocks = limit; 5951689c958cSLi Xi buf->f_bfree = buf->f_bavail = 5952689c958cSLi Xi (buf->f_blocks > curblock) ? 5953689c958cSLi Xi (buf->f_blocks - curblock) : 0; 5954689c958cSLi Xi } 5955689c958cSLi Xi 5956a08fe66eSChengguang Xu limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit, 5957a08fe66eSChengguang Xu dquot->dq_dqb.dqb_ihardlimit); 5958689c958cSLi Xi if (limit && buf->f_files > limit) { 5959689c958cSLi Xi buf->f_files = limit; 5960689c958cSLi Xi buf->f_ffree = 5961689c958cSLi Xi (buf->f_files > dquot->dq_dqb.dqb_curinodes) ? 5962689c958cSLi Xi (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0; 5963689c958cSLi Xi } 5964689c958cSLi Xi 59657b9ca4c6SJan Kara spin_unlock(&dquot->dq_dqb_lock); 5966689c958cSLi Xi dqput(dquot); 5967689c958cSLi Xi return 0; 5968689c958cSLi Xi } 5969689c958cSLi Xi #endif 5970689c958cSLi Xi 5971617ba13bSMingming Cao static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf) 5972ac27a0ecSDave Kleikamp { 5973ac27a0ecSDave Kleikamp struct super_block *sb = dentry->d_sb; 5974617ba13bSMingming Cao struct ext4_sb_info *sbi = EXT4_SB(sb); 5975617ba13bSMingming Cao struct ext4_super_block *es = sbi->s_es; 597627dd4385SLukas Czerner ext4_fsblk_t overhead = 0, resv_blocks; 5977960cc398SPekka Enberg u64 fsid; 5978d02a9391SKazuya Mio s64 bfree; 597927dd4385SLukas Czerner resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters)); 5980ac27a0ecSDave Kleikamp 5981952fc18eSTheodore Ts'o if (!test_opt(sb, MINIX_DF)) 5982952fc18eSTheodore Ts'o overhead = sbi->s_overhead; 5983ac27a0ecSDave Kleikamp 5984617ba13bSMingming Cao buf->f_type = EXT4_SUPER_MAGIC; 5985ac27a0ecSDave Kleikamp buf->f_bsize = sb->s_blocksize; 5986b72f78cbSEric Sandeen buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead); 598757042651STheodore Ts'o bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) - 598857042651STheodore Ts'o percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter); 5989d02a9391SKazuya Mio /* prevent underflow in case that few free space is available */ 599057042651STheodore Ts'o buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0)); 599127dd4385SLukas Czerner buf->f_bavail = buf->f_bfree - 599227dd4385SLukas Czerner (ext4_r_blocks_count(es) + resv_blocks); 599327dd4385SLukas Czerner if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks)) 5994ac27a0ecSDave Kleikamp buf->f_bavail = 0; 5995ac27a0ecSDave Kleikamp buf->f_files = le32_to_cpu(es->s_inodes_count); 599652d9f3b4SPeter Zijlstra buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter); 5997617ba13bSMingming Cao buf->f_namelen = EXT4_NAME_LEN; 5998960cc398SPekka Enberg fsid = le64_to_cpup((void *)es->s_uuid) ^ 5999960cc398SPekka Enberg le64_to_cpup((void *)es->s_uuid + sizeof(u64)); 6000960cc398SPekka Enberg buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL; 6001960cc398SPekka Enberg buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL; 60020b8e58a1SAndreas Dilger 6003689c958cSLi Xi #ifdef CONFIG_QUOTA 6004689c958cSLi Xi if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) && 6005689c958cSLi Xi sb_has_quota_limits_enabled(sb, PRJQUOTA)) 6006689c958cSLi Xi ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf); 6007689c958cSLi Xi #endif 6008ac27a0ecSDave Kleikamp return 0; 6009ac27a0ecSDave Kleikamp } 6010ac27a0ecSDave Kleikamp 6011ac27a0ecSDave Kleikamp 6012ac27a0ecSDave Kleikamp #ifdef CONFIG_QUOTA 6013ac27a0ecSDave Kleikamp 6014bc8230eeSJan Kara /* 6015bc8230eeSJan Kara * Helper functions so that transaction is started before we acquire dqio_sem 6016bc8230eeSJan Kara * to keep correct lock ordering of transaction > dqio_sem 6017bc8230eeSJan Kara */ 6018ac27a0ecSDave Kleikamp static inline struct inode *dquot_to_inode(struct dquot *dquot) 6019ac27a0ecSDave Kleikamp { 60204c376dcaSEric W. Biederman return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type]; 6021ac27a0ecSDave Kleikamp } 6022ac27a0ecSDave Kleikamp 6023617ba13bSMingming Cao static int ext4_write_dquot(struct dquot *dquot) 6024ac27a0ecSDave Kleikamp { 6025ac27a0ecSDave Kleikamp int ret, err; 6026ac27a0ecSDave Kleikamp handle_t *handle; 6027ac27a0ecSDave Kleikamp struct inode *inode; 6028ac27a0ecSDave Kleikamp 6029ac27a0ecSDave Kleikamp inode = dquot_to_inode(dquot); 60309924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 6031617ba13bSMingming Cao EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb)); 6032ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6033ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6034ac27a0ecSDave Kleikamp ret = dquot_commit(dquot); 6035617ba13bSMingming Cao err = ext4_journal_stop(handle); 6036ac27a0ecSDave Kleikamp if (!ret) 6037ac27a0ecSDave Kleikamp ret = err; 6038ac27a0ecSDave Kleikamp return ret; 6039ac27a0ecSDave Kleikamp } 6040ac27a0ecSDave Kleikamp 6041617ba13bSMingming Cao static int ext4_acquire_dquot(struct dquot *dquot) 6042ac27a0ecSDave Kleikamp { 6043ac27a0ecSDave Kleikamp int ret, err; 6044ac27a0ecSDave Kleikamp handle_t *handle; 6045ac27a0ecSDave Kleikamp 60469924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6047617ba13bSMingming Cao EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb)); 6048ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6049ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6050ac27a0ecSDave Kleikamp ret = dquot_acquire(dquot); 6051617ba13bSMingming Cao err = ext4_journal_stop(handle); 6052ac27a0ecSDave Kleikamp if (!ret) 6053ac27a0ecSDave Kleikamp ret = err; 6054ac27a0ecSDave Kleikamp return ret; 6055ac27a0ecSDave Kleikamp } 6056ac27a0ecSDave Kleikamp 6057617ba13bSMingming Cao static int ext4_release_dquot(struct dquot *dquot) 6058ac27a0ecSDave Kleikamp { 6059ac27a0ecSDave Kleikamp int ret, err; 6060ac27a0ecSDave Kleikamp handle_t *handle; 6061ac27a0ecSDave Kleikamp 60629924a92aSTheodore Ts'o handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA, 6063617ba13bSMingming Cao EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb)); 60649c3013e9SJan Kara if (IS_ERR(handle)) { 60659c3013e9SJan Kara /* Release dquot anyway to avoid endless cycle in dqput() */ 60669c3013e9SJan Kara dquot_release(dquot); 6067ac27a0ecSDave Kleikamp return PTR_ERR(handle); 60689c3013e9SJan Kara } 6069ac27a0ecSDave Kleikamp ret = dquot_release(dquot); 6070617ba13bSMingming Cao err = ext4_journal_stop(handle); 6071ac27a0ecSDave Kleikamp if (!ret) 6072ac27a0ecSDave Kleikamp ret = err; 6073ac27a0ecSDave Kleikamp return ret; 6074ac27a0ecSDave Kleikamp } 6075ac27a0ecSDave Kleikamp 6076617ba13bSMingming Cao static int ext4_mark_dquot_dirty(struct dquot *dquot) 6077ac27a0ecSDave Kleikamp { 6078262b4662SJan Kara struct super_block *sb = dquot->dq_sb; 6079262b4662SJan Kara struct ext4_sb_info *sbi = EXT4_SB(sb); 6080262b4662SJan Kara 60812c8be6b2SJan Kara /* Are we journaling quotas? */ 6082e2b911c5SDarrick J. Wong if (ext4_has_feature_quota(sb) || 6083262b4662SJan Kara sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) { 6084ac27a0ecSDave Kleikamp dquot_mark_dquot_dirty(dquot); 6085617ba13bSMingming Cao return ext4_write_dquot(dquot); 6086ac27a0ecSDave Kleikamp } else { 6087ac27a0ecSDave Kleikamp return dquot_mark_dquot_dirty(dquot); 6088ac27a0ecSDave Kleikamp } 6089ac27a0ecSDave Kleikamp } 6090ac27a0ecSDave Kleikamp 6091617ba13bSMingming Cao static int ext4_write_info(struct super_block *sb, int type) 6092ac27a0ecSDave Kleikamp { 6093ac27a0ecSDave Kleikamp int ret, err; 6094ac27a0ecSDave Kleikamp handle_t *handle; 6095ac27a0ecSDave Kleikamp 6096ac27a0ecSDave Kleikamp /* Data block + inode block */ 60972b0143b5SDavid Howells handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2); 6098ac27a0ecSDave Kleikamp if (IS_ERR(handle)) 6099ac27a0ecSDave Kleikamp return PTR_ERR(handle); 6100ac27a0ecSDave Kleikamp ret = dquot_commit_info(sb, type); 6101617ba13bSMingming Cao err = ext4_journal_stop(handle); 6102ac27a0ecSDave Kleikamp if (!ret) 6103ac27a0ecSDave Kleikamp ret = err; 6104ac27a0ecSDave Kleikamp return ret; 6105ac27a0ecSDave Kleikamp } 6106ac27a0ecSDave Kleikamp 6107ac27a0ecSDave Kleikamp /* 6108ac27a0ecSDave Kleikamp * Turn on quotas during mount time - we need to find 6109ac27a0ecSDave Kleikamp * the quota file and such... 6110ac27a0ecSDave Kleikamp */ 6111617ba13bSMingming Cao static int ext4_quota_on_mount(struct super_block *sb, int type) 6112ac27a0ecSDave Kleikamp { 611333458eabSTheodore Ts'o return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type), 6114617ba13bSMingming Cao EXT4_SB(sb)->s_jquota_fmt, type); 6115ac27a0ecSDave Kleikamp } 6116ac27a0ecSDave Kleikamp 6117daf647d2STheodore Ts'o static void lockdep_set_quota_inode(struct inode *inode, int subclass) 6118daf647d2STheodore Ts'o { 6119daf647d2STheodore Ts'o struct ext4_inode_info *ei = EXT4_I(inode); 6120daf647d2STheodore Ts'o 6121daf647d2STheodore Ts'o /* The first argument of lockdep_set_subclass has to be 6122daf647d2STheodore Ts'o * *exactly* the same as the argument to init_rwsem() --- in 6123daf647d2STheodore Ts'o * this case, in init_once() --- or lockdep gets unhappy 6124daf647d2STheodore Ts'o * because the name of the lock is set using the 6125daf647d2STheodore Ts'o * stringification of the argument to init_rwsem(). 6126daf647d2STheodore Ts'o */ 6127daf647d2STheodore Ts'o (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */ 6128daf647d2STheodore Ts'o lockdep_set_subclass(&ei->i_data_sem, subclass); 6129daf647d2STheodore Ts'o } 6130daf647d2STheodore Ts'o 6131ac27a0ecSDave Kleikamp /* 6132ac27a0ecSDave Kleikamp * Standard function to be called on quota_on 6133ac27a0ecSDave Kleikamp */ 6134617ba13bSMingming Cao static int ext4_quota_on(struct super_block *sb, int type, int format_id, 61358c54ca9cSAl Viro const struct path *path) 6136ac27a0ecSDave Kleikamp { 6137ac27a0ecSDave Kleikamp int err; 6138ac27a0ecSDave Kleikamp 6139ac27a0ecSDave Kleikamp if (!test_opt(sb, QUOTA)) 6140ac27a0ecSDave Kleikamp return -EINVAL; 61410623543bSJan Kara 6142ac27a0ecSDave Kleikamp /* Quotafile not on the same filesystem? */ 6143d8c9584eSAl Viro if (path->dentry->d_sb != sb) 6144ac27a0ecSDave Kleikamp return -EXDEV; 61450623543bSJan Kara /* Journaling quota? */ 61460623543bSJan Kara if (EXT4_SB(sb)->s_qf_names[type]) { 61472b2d6d01STheodore Ts'o /* Quotafile not in fs root? */ 6148f00c9e44SJan Kara if (path->dentry->d_parent != sb->s_root) 6149b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, 6150b31e1552SEric Sandeen "Quota file not on filesystem root. " 6151b31e1552SEric Sandeen "Journaled quota will not work"); 615291389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY; 615391389240SJan Kara } else { 615491389240SJan Kara /* 615591389240SJan Kara * Clear the flag just in case mount options changed since 615691389240SJan Kara * last time. 615791389240SJan Kara */ 615891389240SJan Kara sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY; 61590623543bSJan Kara } 61600623543bSJan Kara 61610623543bSJan Kara /* 61620623543bSJan Kara * When we journal data on quota file, we have to flush journal to see 61630623543bSJan Kara * all updates to the file when we bypass pagecache... 61640623543bSJan Kara */ 61650390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && 61662b0143b5SDavid Howells ext4_should_journal_data(d_inode(path->dentry))) { 61670623543bSJan Kara /* 61680623543bSJan Kara * We don't need to lock updates but journal_flush() could 61690623543bSJan Kara * otherwise be livelocked... 61700623543bSJan Kara */ 61710623543bSJan Kara jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal); 61727ffe1ea8SHidehiro Kawai err = jbd2_journal_flush(EXT4_SB(sb)->s_journal); 61730623543bSJan Kara jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal); 6174f00c9e44SJan Kara if (err) 61757ffe1ea8SHidehiro Kawai return err; 61767ffe1ea8SHidehiro Kawai } 6177957153fcSJan Kara 6178daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA); 6179daf647d2STheodore Ts'o err = dquot_quota_on(sb, type, format_id, path); 6180957153fcSJan Kara if (err) { 6181daf647d2STheodore Ts'o lockdep_set_quota_inode(path->dentry->d_inode, 6182daf647d2STheodore Ts'o I_DATA_SEM_NORMAL); 6183957153fcSJan Kara } else { 6184957153fcSJan Kara struct inode *inode = d_inode(path->dentry); 6185957153fcSJan Kara handle_t *handle; 6186957153fcSJan Kara 618761a92987SJan Kara /* 618861a92987SJan Kara * Set inode flags to prevent userspace from messing with quota 618961a92987SJan Kara * files. If this fails, we return success anyway since quotas 619061a92987SJan Kara * are already enabled and this is not a hard failure. 619161a92987SJan Kara */ 6192957153fcSJan Kara inode_lock(inode); 6193957153fcSJan Kara handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 6194957153fcSJan Kara if (IS_ERR(handle)) 6195957153fcSJan Kara goto unlock_inode; 6196957153fcSJan Kara EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL; 6197957153fcSJan Kara inode_set_flags(inode, S_NOATIME | S_IMMUTABLE, 6198957153fcSJan Kara S_NOATIME | S_IMMUTABLE); 61994209ae12SHarshad Shirwadkar err = ext4_mark_inode_dirty(handle, inode); 6200957153fcSJan Kara ext4_journal_stop(handle); 6201957153fcSJan Kara unlock_inode: 6202957153fcSJan Kara inode_unlock(inode); 6203957153fcSJan Kara } 6204daf647d2STheodore Ts'o return err; 6205ac27a0ecSDave Kleikamp } 6206ac27a0ecSDave Kleikamp 62077c319d32SAditya Kali static int ext4_quota_enable(struct super_block *sb, int type, int format_id, 62087c319d32SAditya Kali unsigned int flags) 62097c319d32SAditya Kali { 62107c319d32SAditya Kali int err; 62117c319d32SAditya Kali struct inode *qf_inode; 6212a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 62137c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6214689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6215689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 62167c319d32SAditya Kali }; 62177c319d32SAditya Kali 6218e2b911c5SDarrick J. Wong BUG_ON(!ext4_has_feature_quota(sb)); 62197c319d32SAditya Kali 62207c319d32SAditya Kali if (!qf_inums[type]) 62217c319d32SAditya Kali return -EPERM; 62227c319d32SAditya Kali 62238a363970STheodore Ts'o qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL); 62247c319d32SAditya Kali if (IS_ERR(qf_inode)) { 62257c319d32SAditya Kali ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]); 62267c319d32SAditya Kali return PTR_ERR(qf_inode); 62277c319d32SAditya Kali } 62287c319d32SAditya Kali 6229bcb13850SJan Kara /* Don't account quota for quota files to avoid recursion */ 6230bcb13850SJan Kara qf_inode->i_flags |= S_NOQUOTA; 6231daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA); 62327212b95eSJan Kara err = dquot_load_quota_inode(qf_inode, type, format_id, flags); 6233daf647d2STheodore Ts'o if (err) 6234daf647d2STheodore Ts'o lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL); 623561157b24SPan Bian iput(qf_inode); 62367c319d32SAditya Kali 62377c319d32SAditya Kali return err; 62387c319d32SAditya Kali } 62397c319d32SAditya Kali 62407c319d32SAditya Kali /* Enable usage tracking for all quota types. */ 62417c319d32SAditya Kali static int ext4_enable_quotas(struct super_block *sb) 62427c319d32SAditya Kali { 62437c319d32SAditya Kali int type, err = 0; 6244a2d4a646SJan Kara unsigned long qf_inums[EXT4_MAXQUOTAS] = { 62457c319d32SAditya Kali le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum), 6246689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum), 6247689c958cSLi Xi le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum) 62487c319d32SAditya Kali }; 624949da9392SJan Kara bool quota_mopt[EXT4_MAXQUOTAS] = { 625049da9392SJan Kara test_opt(sb, USRQUOTA), 625149da9392SJan Kara test_opt(sb, GRPQUOTA), 625249da9392SJan Kara test_opt(sb, PRJQUOTA), 625349da9392SJan Kara }; 62547c319d32SAditya Kali 625591389240SJan Kara sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY; 6256a2d4a646SJan Kara for (type = 0; type < EXT4_MAXQUOTAS; type++) { 62577c319d32SAditya Kali if (qf_inums[type]) { 62587c319d32SAditya Kali err = ext4_quota_enable(sb, type, QFMT_VFS_V1, 625949da9392SJan Kara DQUOT_USAGE_ENABLED | 626049da9392SJan Kara (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0)); 62617c319d32SAditya Kali if (err) { 62627c319d32SAditya Kali ext4_warning(sb, 626372ba7450STheodore Ts'o "Failed to enable quota tracking " 626472ba7450STheodore Ts'o "(type=%d, err=%d). Please run " 626572ba7450STheodore Ts'o "e2fsck to fix.", type, err); 62667f144fd0SJunichi Uekawa for (type--; type >= 0; type--) 62677f144fd0SJunichi Uekawa dquot_quota_off(sb, type); 62687f144fd0SJunichi Uekawa 62697c319d32SAditya Kali return err; 62707c319d32SAditya Kali } 62717c319d32SAditya Kali } 62727c319d32SAditya Kali } 62737c319d32SAditya Kali return 0; 62747c319d32SAditya Kali } 62757c319d32SAditya Kali 6276ca0e05e4SDmitry Monakhov static int ext4_quota_off(struct super_block *sb, int type) 6277ca0e05e4SDmitry Monakhov { 627821f97697SJan Kara struct inode *inode = sb_dqopt(sb)->files[type]; 627921f97697SJan Kara handle_t *handle; 6280957153fcSJan Kara int err; 628121f97697SJan Kara 628287009d86SDmitry Monakhov /* Force all delayed allocation blocks to be allocated. 628387009d86SDmitry Monakhov * Caller already holds s_umount sem */ 628487009d86SDmitry Monakhov if (test_opt(sb, DELALLOC)) 6285ca0e05e4SDmitry Monakhov sync_filesystem(sb); 6286ca0e05e4SDmitry Monakhov 6287957153fcSJan Kara if (!inode || !igrab(inode)) 62880b268590SAmir Goldstein goto out; 62890b268590SAmir Goldstein 6290957153fcSJan Kara err = dquot_quota_off(sb, type); 6291964edf66SJan Kara if (err || ext4_has_feature_quota(sb)) 6292957153fcSJan Kara goto out_put; 6293957153fcSJan Kara 6294957153fcSJan Kara inode_lock(inode); 629561a92987SJan Kara /* 629661a92987SJan Kara * Update modification times of quota files when userspace can 629761a92987SJan Kara * start looking at them. If we fail, we return success anyway since 629861a92987SJan Kara * this is not a hard failure and quotas are already disabled. 629961a92987SJan Kara */ 63009924a92aSTheodore Ts'o handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1); 63014209ae12SHarshad Shirwadkar if (IS_ERR(handle)) { 63024209ae12SHarshad Shirwadkar err = PTR_ERR(handle); 6303957153fcSJan Kara goto out_unlock; 63044209ae12SHarshad Shirwadkar } 6305957153fcSJan Kara EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL); 6306957153fcSJan Kara inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE); 6307eeca7ea1SDeepa Dinamani inode->i_mtime = inode->i_ctime = current_time(inode); 63084209ae12SHarshad Shirwadkar err = ext4_mark_inode_dirty(handle, inode); 630921f97697SJan Kara ext4_journal_stop(handle); 6310957153fcSJan Kara out_unlock: 6311957153fcSJan Kara inode_unlock(inode); 6312957153fcSJan Kara out_put: 6313964edf66SJan Kara lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL); 6314957153fcSJan Kara iput(inode); 6315957153fcSJan Kara return err; 631621f97697SJan Kara out: 6317ca0e05e4SDmitry Monakhov return dquot_quota_off(sb, type); 6318ca0e05e4SDmitry Monakhov } 6319ca0e05e4SDmitry Monakhov 6320ac27a0ecSDave Kleikamp /* Read data from quotafile - avoid pagecache and such because we cannot afford 6321ac27a0ecSDave Kleikamp * acquiring the locks... As quota files are never truncated and quota code 6322ac27a0ecSDave Kleikamp * itself serializes the operations (and no one else should touch the files) 6323ac27a0ecSDave Kleikamp * we don't have to be afraid of races */ 6324617ba13bSMingming Cao static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 6325ac27a0ecSDave Kleikamp size_t len, loff_t off) 6326ac27a0ecSDave Kleikamp { 6327ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6328725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 6329ac27a0ecSDave Kleikamp int offset = off & (sb->s_blocksize - 1); 6330ac27a0ecSDave Kleikamp int tocopy; 6331ac27a0ecSDave Kleikamp size_t toread; 6332ac27a0ecSDave Kleikamp struct buffer_head *bh; 6333ac27a0ecSDave Kleikamp loff_t i_size = i_size_read(inode); 6334ac27a0ecSDave Kleikamp 6335ac27a0ecSDave Kleikamp if (off > i_size) 6336ac27a0ecSDave Kleikamp return 0; 6337ac27a0ecSDave Kleikamp if (off+len > i_size) 6338ac27a0ecSDave Kleikamp len = i_size-off; 6339ac27a0ecSDave Kleikamp toread = len; 6340ac27a0ecSDave Kleikamp while (toread > 0) { 6341ac27a0ecSDave Kleikamp tocopy = sb->s_blocksize - offset < toread ? 6342ac27a0ecSDave Kleikamp sb->s_blocksize - offset : toread; 63431c215028STheodore Ts'o bh = ext4_bread(NULL, inode, blk, 0); 63441c215028STheodore Ts'o if (IS_ERR(bh)) 63451c215028STheodore Ts'o return PTR_ERR(bh); 6346ac27a0ecSDave Kleikamp if (!bh) /* A hole? */ 6347ac27a0ecSDave Kleikamp memset(data, 0, tocopy); 6348ac27a0ecSDave Kleikamp else 6349ac27a0ecSDave Kleikamp memcpy(data, bh->b_data+offset, tocopy); 6350ac27a0ecSDave Kleikamp brelse(bh); 6351ac27a0ecSDave Kleikamp offset = 0; 6352ac27a0ecSDave Kleikamp toread -= tocopy; 6353ac27a0ecSDave Kleikamp data += tocopy; 6354ac27a0ecSDave Kleikamp blk++; 6355ac27a0ecSDave Kleikamp } 6356ac27a0ecSDave Kleikamp return len; 6357ac27a0ecSDave Kleikamp } 6358ac27a0ecSDave Kleikamp 6359ac27a0ecSDave Kleikamp /* Write to quotafile (we know the transaction is already started and has 6360ac27a0ecSDave Kleikamp * enough credits) */ 6361617ba13bSMingming Cao static ssize_t ext4_quota_write(struct super_block *sb, int type, 6362ac27a0ecSDave Kleikamp const char *data, size_t len, loff_t off) 6363ac27a0ecSDave Kleikamp { 6364ac27a0ecSDave Kleikamp struct inode *inode = sb_dqopt(sb)->files[type]; 6365725d26d3SAneesh Kumar K.V ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 63664209ae12SHarshad Shirwadkar int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1); 6367c5e298aeSTheodore Ts'o int retries = 0; 6368ac27a0ecSDave Kleikamp struct buffer_head *bh; 6369ac27a0ecSDave Kleikamp handle_t *handle = journal_current_handle(); 6370ac27a0ecSDave Kleikamp 63710390131bSFrank Mayhar if (EXT4_SB(sb)->s_journal && !handle) { 6372b31e1552SEric Sandeen ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 6373b31e1552SEric Sandeen " cancelled because transaction is not started", 63749c3013e9SJan Kara (unsigned long long)off, (unsigned long long)len); 63759c3013e9SJan Kara return -EIO; 63769c3013e9SJan Kara } 637767eeb568SDmitry Monakhov /* 637867eeb568SDmitry Monakhov * Since we account only one data block in transaction credits, 637967eeb568SDmitry Monakhov * then it is impossible to cross a block boundary. 638067eeb568SDmitry Monakhov */ 638167eeb568SDmitry Monakhov if (sb->s_blocksize - offset < len) { 638267eeb568SDmitry Monakhov ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)" 638367eeb568SDmitry Monakhov " cancelled because not block aligned", 638467eeb568SDmitry Monakhov (unsigned long long)off, (unsigned long long)len); 638567eeb568SDmitry Monakhov return -EIO; 638667eeb568SDmitry Monakhov } 638767eeb568SDmitry Monakhov 6388c5e298aeSTheodore Ts'o do { 6389c5e298aeSTheodore Ts'o bh = ext4_bread(handle, inode, blk, 6390c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_CREATE | 6391c5e298aeSTheodore Ts'o EXT4_GET_BLOCKS_METADATA_NOFAIL); 639245586c70SMasahiro Yamada } while (PTR_ERR(bh) == -ENOSPC && 6393c5e298aeSTheodore Ts'o ext4_should_retry_alloc(inode->i_sb, &retries)); 63941c215028STheodore Ts'o if (IS_ERR(bh)) 63951c215028STheodore Ts'o return PTR_ERR(bh); 6396ac27a0ecSDave Kleikamp if (!bh) 6397ac27a0ecSDave Kleikamp goto out; 63985d601255Sliang xie BUFFER_TRACE(bh, "get write access"); 6399617ba13bSMingming Cao err = ext4_journal_get_write_access(handle, bh); 6400ac27a0ecSDave Kleikamp if (err) { 6401ac27a0ecSDave Kleikamp brelse(bh); 64021c215028STheodore Ts'o return err; 6403ac27a0ecSDave Kleikamp } 6404ac27a0ecSDave Kleikamp lock_buffer(bh); 640567eeb568SDmitry Monakhov memcpy(bh->b_data+offset, data, len); 6406ac27a0ecSDave Kleikamp flush_dcache_page(bh->b_page); 6407ac27a0ecSDave Kleikamp unlock_buffer(bh); 64080390131bSFrank Mayhar err = ext4_handle_dirty_metadata(handle, NULL, bh); 6409ac27a0ecSDave Kleikamp brelse(bh); 6410ac27a0ecSDave Kleikamp out: 641167eeb568SDmitry Monakhov if (inode->i_size < off + len) { 641267eeb568SDmitry Monakhov i_size_write(inode, off + len); 6413617ba13bSMingming Cao EXT4_I(inode)->i_disksize = inode->i_size; 64144209ae12SHarshad Shirwadkar err2 = ext4_mark_inode_dirty(handle, inode); 64154209ae12SHarshad Shirwadkar if (unlikely(err2 && !err)) 64164209ae12SHarshad Shirwadkar err = err2; 641721f97697SJan Kara } 64184209ae12SHarshad Shirwadkar return err ? err : len; 6419ac27a0ecSDave Kleikamp } 6420ac27a0ecSDave Kleikamp #endif 6421ac27a0ecSDave Kleikamp 6422152a0836SAl Viro static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags, 6423152a0836SAl Viro const char *dev_name, void *data) 6424ac27a0ecSDave Kleikamp { 6425152a0836SAl Viro return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super); 6426ac27a0ecSDave Kleikamp } 6427ac27a0ecSDave Kleikamp 6428c290ea01SJan Kara #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2) 642924b58424STheodore Ts'o static inline void register_as_ext2(void) 643024b58424STheodore Ts'o { 643124b58424STheodore Ts'o int err = register_filesystem(&ext2_fs_type); 643224b58424STheodore Ts'o if (err) 643324b58424STheodore Ts'o printk(KERN_WARNING 643424b58424STheodore Ts'o "EXT4-fs: Unable to register as ext2 (%d)\n", err); 643524b58424STheodore Ts'o } 643624b58424STheodore Ts'o 643724b58424STheodore Ts'o static inline void unregister_as_ext2(void) 643824b58424STheodore Ts'o { 643924b58424STheodore Ts'o unregister_filesystem(&ext2_fs_type); 644024b58424STheodore Ts'o } 64412035e776STheodore Ts'o 64422035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) 64432035e776STheodore Ts'o { 6444e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_incompat_features(sb)) 64452035e776STheodore Ts'o return 0; 6446bc98a42cSDavid Howells if (sb_rdonly(sb)) 64472035e776STheodore Ts'o return 1; 6448e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext2_ro_compat_features(sb)) 64492035e776STheodore Ts'o return 0; 64502035e776STheodore Ts'o return 1; 64512035e776STheodore Ts'o } 645224b58424STheodore Ts'o #else 645324b58424STheodore Ts'o static inline void register_as_ext2(void) { } 645424b58424STheodore Ts'o static inline void unregister_as_ext2(void) { } 64552035e776STheodore Ts'o static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; } 645624b58424STheodore Ts'o #endif 645724b58424STheodore Ts'o 645824b58424STheodore Ts'o static inline void register_as_ext3(void) 645924b58424STheodore Ts'o { 646024b58424STheodore Ts'o int err = register_filesystem(&ext3_fs_type); 646124b58424STheodore Ts'o if (err) 646224b58424STheodore Ts'o printk(KERN_WARNING 646324b58424STheodore Ts'o "EXT4-fs: Unable to register as ext3 (%d)\n", err); 646424b58424STheodore Ts'o } 646524b58424STheodore Ts'o 646624b58424STheodore Ts'o static inline void unregister_as_ext3(void) 646724b58424STheodore Ts'o { 646824b58424STheodore Ts'o unregister_filesystem(&ext3_fs_type); 646924b58424STheodore Ts'o } 64702035e776STheodore Ts'o 64712035e776STheodore Ts'o static inline int ext3_feature_set_ok(struct super_block *sb) 64722035e776STheodore Ts'o { 6473e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_incompat_features(sb)) 64742035e776STheodore Ts'o return 0; 6475e2b911c5SDarrick J. Wong if (!ext4_has_feature_journal(sb)) 64762035e776STheodore Ts'o return 0; 6477bc98a42cSDavid Howells if (sb_rdonly(sb)) 64782035e776STheodore Ts'o return 1; 6479e2b911c5SDarrick J. Wong if (ext4_has_unknown_ext3_ro_compat_features(sb)) 64802035e776STheodore Ts'o return 0; 64812035e776STheodore Ts'o return 1; 64822035e776STheodore Ts'o } 648324b58424STheodore Ts'o 648403010a33STheodore Ts'o static struct file_system_type ext4_fs_type = { 6485ac27a0ecSDave Kleikamp .owner = THIS_MODULE, 648603010a33STheodore Ts'o .name = "ext4", 6487152a0836SAl Viro .mount = ext4_mount, 6488ac27a0ecSDave Kleikamp .kill_sb = kill_block_super, 6489ac27a0ecSDave Kleikamp .fs_flags = FS_REQUIRES_DEV, 6490ac27a0ecSDave Kleikamp }; 64917f78e035SEric W. Biederman MODULE_ALIAS_FS("ext4"); 6492ac27a0ecSDave Kleikamp 6493e9e3bcecSEric Sandeen /* Shared across all ext4 file systems */ 6494e9e3bcecSEric Sandeen wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ]; 6495e9e3bcecSEric Sandeen 64965dabfc78STheodore Ts'o static int __init ext4_init_fs(void) 6497ac27a0ecSDave Kleikamp { 6498e9e3bcecSEric Sandeen int i, err; 6499c9de560dSAlex Tomas 6500e294a537STheodore Ts'o ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64); 650107c0c5d8SAl Viro ext4_li_info = NULL; 650207c0c5d8SAl Viro mutex_init(&ext4_li_mtx); 650307c0c5d8SAl Viro 65049a4c8019SCarlos Maiolino /* Build-time check for flags consistency */ 650512e9b892SDmitry Monakhov ext4_check_flag_values(); 6506e9e3bcecSEric Sandeen 6507e142d052SJan Kara for (i = 0; i < EXT4_WQ_HASH_SZ; i++) 6508e9e3bcecSEric Sandeen init_waitqueue_head(&ext4__ioend_wq[i]); 6509e9e3bcecSEric Sandeen 651051865fdaSZheng Liu err = ext4_init_es(); 65116fd058f7STheodore Ts'o if (err) 65126fd058f7STheodore Ts'o return err; 651351865fdaSZheng Liu 65141dc0aa46SEric Whitney err = ext4_init_pending(); 65151dc0aa46SEric Whitney if (err) 651622cfe4b4SEric Biggers goto out7; 651722cfe4b4SEric Biggers 651822cfe4b4SEric Biggers err = ext4_init_post_read_processing(); 651922cfe4b4SEric Biggers if (err) 65201dc0aa46SEric Whitney goto out6; 65211dc0aa46SEric Whitney 652251865fdaSZheng Liu err = ext4_init_pageio(); 652351865fdaSZheng Liu if (err) 6524b5799018STheodore Ts'o goto out5; 652551865fdaSZheng Liu 65265dabfc78STheodore Ts'o err = ext4_init_system_zone(); 6527bd2d0210STheodore Ts'o if (err) 6528b5799018STheodore Ts'o goto out4; 6529857ac889SLukas Czerner 6530b5799018STheodore Ts'o err = ext4_init_sysfs(); 6531dd68314cSTheodore Ts'o if (err) 6532b5799018STheodore Ts'o goto out3; 6533857ac889SLukas Czerner 65345dabfc78STheodore Ts'o err = ext4_init_mballoc(); 6535ac27a0ecSDave Kleikamp if (err) 6536c9de560dSAlex Tomas goto out2; 6537ac27a0ecSDave Kleikamp err = init_inodecache(); 6538ac27a0ecSDave Kleikamp if (err) 6539ac27a0ecSDave Kleikamp goto out1; 654024b58424STheodore Ts'o register_as_ext3(); 65412035e776STheodore Ts'o register_as_ext2(); 654203010a33STheodore Ts'o err = register_filesystem(&ext4_fs_type); 6543ac27a0ecSDave Kleikamp if (err) 6544ac27a0ecSDave Kleikamp goto out; 6545bfff6873SLukas Czerner 6546ac27a0ecSDave Kleikamp return 0; 6547ac27a0ecSDave Kleikamp out: 654824b58424STheodore Ts'o unregister_as_ext2(); 654924b58424STheodore Ts'o unregister_as_ext3(); 6550ac27a0ecSDave Kleikamp destroy_inodecache(); 6551ac27a0ecSDave Kleikamp out1: 65525dabfc78STheodore Ts'o ext4_exit_mballoc(); 65539c191f70ST Makphaibulchoke out2: 6554b5799018STheodore Ts'o ext4_exit_sysfs(); 6555b5799018STheodore Ts'o out3: 6556dd68314cSTheodore Ts'o ext4_exit_system_zone(); 6557b5799018STheodore Ts'o out4: 65585dabfc78STheodore Ts'o ext4_exit_pageio(); 6559b5799018STheodore Ts'o out5: 656022cfe4b4SEric Biggers ext4_exit_post_read_processing(); 65611dc0aa46SEric Whitney out6: 656222cfe4b4SEric Biggers ext4_exit_pending(); 656322cfe4b4SEric Biggers out7: 656451865fdaSZheng Liu ext4_exit_es(); 656551865fdaSZheng Liu 6566ac27a0ecSDave Kleikamp return err; 6567ac27a0ecSDave Kleikamp } 6568ac27a0ecSDave Kleikamp 65695dabfc78STheodore Ts'o static void __exit ext4_exit_fs(void) 6570ac27a0ecSDave Kleikamp { 6571bfff6873SLukas Czerner ext4_destroy_lazyinit_thread(); 657224b58424STheodore Ts'o unregister_as_ext2(); 657324b58424STheodore Ts'o unregister_as_ext3(); 657403010a33STheodore Ts'o unregister_filesystem(&ext4_fs_type); 6575ac27a0ecSDave Kleikamp destroy_inodecache(); 65765dabfc78STheodore Ts'o ext4_exit_mballoc(); 6577b5799018STheodore Ts'o ext4_exit_sysfs(); 65785dabfc78STheodore Ts'o ext4_exit_system_zone(); 65795dabfc78STheodore Ts'o ext4_exit_pageio(); 658022cfe4b4SEric Biggers ext4_exit_post_read_processing(); 6581dd12ed14SEric Sandeen ext4_exit_es(); 65821dc0aa46SEric Whitney ext4_exit_pending(); 6583ac27a0ecSDave Kleikamp } 6584ac27a0ecSDave Kleikamp 6585ac27a0ecSDave Kleikamp MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 658683982b6fSTheodore Ts'o MODULE_DESCRIPTION("Fourth Extended Filesystem"); 6587ac27a0ecSDave Kleikamp MODULE_LICENSE("GPL"); 65887ef79ad5STheodore Ts'o MODULE_SOFTDEP("pre: crc32c"); 65895dabfc78STheodore Ts'o module_init(ext4_init_fs) 65905dabfc78STheodore Ts'o module_exit(ext4_exit_fs) 6591